aes: specify the required OpenSSL version

The code in the aes elements assumes OpenSSL >= 1.1.0:

  - implicit library initialization;
  - version retrieved with OpenSSL_version(OPENSSL_VERSION);

and it fails to build with older versions.

Specify the required OpenSSL version explicitly in meson.build so that
the elements are excluded on older systems (e.g. Ubuntu 16.04) and the
rest of GStreamer can still build.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1067>

ci: fix plugins cache diffs directory

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/867>

audio/video aggregator: make use of new aggregator inactive pad API

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/867>

aggregator: expose API for ignoring inactive pads

An inactive pad is a pad which, in live mode, hasn't yet received
a first buffer, but has been waited on at least once.

Exposing API to support this behaviour allows users of aggregator
subclasses to request pads, and not start pushing data on those
immediately, while avoiding systematic timeouts.

Subclasses must check in explicitly to this behavior, most likely
by exposing a user-facing property, and must check whether a pad
needs ignoring when aggregating. That is because by design,
aggregator subclasses don't get a list of "ready" pads, but instead
directly iterate element->sinkpads.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/867>

docs: tutorial 1: error out if an error occurred in Hello World tutorial

Very confusing if it just exits silently with exit code 0 on error.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1185>

discoverer: Advertise container-specific tags with a new API

Since commit a55dafe341ac7398e7c37c30d8b760228296da92, stream-scoped tags no
longer appeared as top-level tags, introducing a behaviour regression, specially
for MP3 files.

The `gst_discoverer_info_get_tags()` API now returns all tags detected for the
given media, as documented.

A new API is introduced to get container-specific tags,
`gst_discoverer_container_info_get_tags()`. The discoverer tool was adapted to
use it. `gst_discoverer_info_get_tags()` is now deprecated in favor of
`gst_discoverer_container_info_get_tags()` and
`gst_discoverer_stream_info_get_tags()`.

Fixes #759

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1107>

vadeinterlace: Accept ANY feature.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1024>

vadeinterlace: Fixate interlace-mode and framerate accordingly.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1024>

vapostproc: Accept ANY feature.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1024>

vapostproc: Traverse caps features in gst_va_vpp_caps_remove_fields()

The previous code had a potential failure for multiple caps features. Now
each caps feature in each structure is reviewed, and if it has a supported
feature, the structure is processed.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1024>

vapostproc: Refactor gst_va_vpp_complete_caps_features()

gst_va_vpp_complete_caps_features() now receives the @feature_name to
add and return if @caps doesn't provide it.

So, instead of two nested loops, now the function is a single loop,
traversing @caps to find if each structure already contains the requested
@features_name.

It's important to add missing caps features with @caps, in order to
not lost information.

The function caller does the external loop by calling per each
available caps feature.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1024>

vapostproc: Split caps transform in two phases.

In order to make more readable the caps transformation, the operation
was split in two phases:

1. Rangify the supported caps structures.
2. Add the missing (and supported) caps features.

Step 1 modified its logic, by copying any unrecognized structure.
It's a previous step required for allowing ANY caps feature as
passthrough.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1024>

tsdemux: Change many GST_{DEBUG, LOG, etc} into _OBJECT

Log files with several demuxers running at once can otherwise get
confusing

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1096>

tsdemux: Issue GST_ELEMENT_WARNING for continuity errors

The application might want to make use of these.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1096>

ci: install newer meson via pip in documentation job

Since updating the existing image seems difficult and efforts
to redo the setup are still work in progress.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1084>

meson: bump meson requirement to >= 0.59

For monorepo build and ugly/bad, for advanced feature
option API like get_option('xyz').required(..) which
we use in combination with the 'gpl' option.

For rest of modules for consistency (people will likely
use newer features based on the top-level requirement).

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1084>

meson: add 'gpl' option and only build plugins with (A)GPL deps if explicitly enabled

Require explicit opt-in to build plugins with (A)GPL dependencies.

Keep ugly/bad options on 'auto' for now so cerbero doesn't fail.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1084>

basetsmux: Support for caps changes

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/981>

basetsmux: Clean up gst_base_ts_mux_create_stream

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/981>

mpegtsbase: Search SCTE-35 DRF_ID_CUEI in multiple registration descriptors

There are streams in the wild that have to add a SCTE-35 trigger in
another e.g. GA94 stream. Most encoders would replace the GA94
descriptor ID with the CUEI one temporarily, but there are some that
will add two registration ID descriptors, one with GA94 and one with
CUEI.

Failing to parse the CUEI registration ID in that case would return
FALSE in _stream_is_private_section , therefore setting it as known PES
and pushing packets downstream instead of calling handle_psi.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/979>

smartencoder: Force reencoding of vp9 for profiles can't support

If the input file is in a profile we won't be able to output (1 or 3 in BGR
format), simply reencode everything for now.

Once we add support for that case we can remove that code

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1062>

encoding-profile: ignore more fields when creating for discover info

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1062>

smartencoder: Always plug a vp9parse when encoding vp9

This way we know that all relevant fields about video format are
added and we avoid renegotiation issues.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1062>

qtmux: Allow more fields changes renegotiating h264

And consider interlace-mode=progress as equivalent to the field not
being specified.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1062>

matroskamux: Ignore some fields when renegotiating

Those values are never used in the muxer so we should not fail the
negotiation if those are changing.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1062>

smartencoder: Fix renegotiating when reencoding parts of the stream with vpx

In the encoded streams we might not have all the information about the
raw video stream, but when reencoding they end up being specified, even
if those are default values.

As vp8 decoders always output frames in some YUV color space we can
ensure that when upstream doesn't specify any value in its caps we
use the default one which is what we end up doing when decoding/reencoding
anyway, so this way downstream (matroskamux in that case) doesn't need
to be able to renegotiate (which it doesn't).

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1062>

tsdemux: Improve gap detection

We should also take into account whether data is currently pending when checking
for gap on streams. It could very well be that some streams have very low
bitrate (and spread out) data. For those we don't want to push out a gap event.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1179>

tsdemux: Handle "negative" timestamps

This is only enabled in push time mode. Furthermore it's only enabled for now if
PCR is to be ignored.

The problem is dealing with streams where the initial PTS/DTS observation might
be greater than following ones (from other PID for example). Before this patch,
this would result in sending buffers without any timestamp which would cause a
wide variety of issues.

Instead, pad segment and buffer timestamps with an extra
value (packetizer->extra_shift, default to 2s), to ensure that we can get valid
timestamps on outgoing buffers (even if that means they are before the segment
start).

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1179>

tsdemux: Handle streams with bogus PTS vs DTS

PTS and DTS should be within a reasonable distance of each other.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1179>

tsdemux: Handle PTS->TS at wraparound

This has been a FIXME for ages.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1179>

webrtcbin: fix prevention of webrtcbin deletion due to ref held by probe callback

https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/810

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1150>

d3d11: d3d11{convert,scale}: Add add-borders property

Functionally identical to that of videoscale element.

When disabled, d3d11convert or d3d11scale element will scale
image without adding borders, meaning that display aspect ratio
will not be preserved.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1170>

vapostproc: Add add-borders property to keep dar

Just as videoscale, it enables add-borders property (FALSE by default)
in vapostproc to add border, if necessary, to keep the display aspect
ratio from the original image.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1169>

vapostproc: Fix early fixation.

First copy missing fields and then fixate all remaining fields.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1169>

ofa: remove ofa audio fingerprinting plugin

I think the MusicIP database for this has been defunct for years,
so I can't imagine this plugin is particularly useful or still
used by anyone.

See https://musicbrainz.org/doc/Fingerprinting#PUID

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1153>

meson: Define cpp_std in the top-level project()

Due to a bug, meson ignores ${lang}_std settings in default_options
for subprojects: https://github.com/mesonbuild/meson/issues/1889

This causes build failures when a subproject requires c++11 or c++14,
etc. Compilers that support those cpp_stds are very common, and all
the toolchains that we support include c++ compilers, so we can
add cpp_std=c++14 to the top-level.

This fixes the webrtc-audio-processing build on Linux, and harfbuzz on
macOS.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1133>

meson: wasapi2,mediafoundation: Work around Windows SDK header issue

Some SDK headers are not standard compliant, so MSVC will
complain when such headers are in use with "/permissive-" compile
option. Use "/Zc:twoPhase-" to work around the issue as documented in
https://docs.microsoft.com/en-us/cpp/build/reference/permissive-standards-conformance?view=msvc-160#windows-header-issues

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1174>

lame:meson: Avoid using fallback when we could use system install

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1176>

subprojects: update vorbis wrap to 1.3.7 and add fallback url

The Windows CI runners complain about self-signed certificates
with the download.xiph.org urls.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1175>

subprojects: update ogg wrap to 1.3.5 and add fallback url

The windows ci runners complain about a self-signed certificate
with the download.xiph.org urls.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1175>

validate:known_issues:rtsp: Mark error message on bus as known

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/814

And reindent file.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1171>

meson: Mark newly fdkaac/ogg/vorbis as allow fallback

This way when the dep is `auto` we will fallback if the system
dependency is not available.

And use https to get libvorbis

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1171>

mediafoundation: Use GetProcAddress() for OS version dependent symbols

We are using some symbols which are not available on Windows 7,
specifically D3D11 interop related ones

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1167>

vapostproc: Transfer colorimetry at fixate if possible.

Taken from videoconvert element.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1110>

meson: Fix warning building GstGLEGL

We forgot to add the `GstGL` gir as a dependency which was properly
done for GstGLWayland and GstGLX11.

And document why we use a list for `all_libraries`

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1168>

meson: Support building lame as subproject

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1068>

meson: Add wrap for libvorbis and libogg

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1068>

meson: Add wrap for fdkaac

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1068>

python:meson: Do not build plugin when statically built

It is not supported yet

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

python: Add an option to disable tests

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

ci: Build introspection support on static builds

But disable python in GES as it is not supported on static builds

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

ges:docs: Ignore GESImageSource as it is plain useless now

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

meson: Add info about 'gstreamer-full'

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

meson: Streamline the way we detect when to build documentation

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

tests: Simple test to use gst-full in python

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

validate:build: Disable static building validate tracer

It introduce duplication of symbols and we need to rework the way we implement it

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

build: Expose ges symbols in gst-full

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

ges:build: Disable python modules when statically building

Introspection fails otherwise

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

Use new `libraries` subprojects variables to statically link then

And also build the girs against `gstreamer-full` when possible. Also
making `gst_init_static_plugins` available from the bindings.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

meson: List libraries and their corresponding gir definition

Introduces a `libraries` variable that contains all libraries in a
list with the following format:

``` meson
libraries = [
    [pkg_name, {
        'lib': library_object
        'gir': [ {full gir definition in a dict } ]
    ],
    ....
]
```

It therefore refactors the way we build the gir so that we can reuse the
same information to build them against 'gstreamer-full' in gst-build
when linking statically

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

meson: Mark files as files()

Making it more robust and future proof

And fix issues that it creates

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

bad:audio: Add generated files sources in declare_dependency

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1093>

python: Fix using overrides when not building PyGObject

Since 547570cd790f2b2e390edc1dfb5df4c7a33de45c we do not always build
PyGObject and our development environment is broken when trying to use
GStreamer python when built against system PyGObject with the following
error importing Gst in there:

```
12345678** (gst-plugin-scanner:710617): CRITICAL **: 11:45:02.343: can't find gi.repository.Gst
Traceback (most recent call last):
  File "/usr/lib/python3.9/site-packages/gi/repository/__init__.py", line 23, in <module>
    from ..importer import DynamicImporter
  File "/usr/lib64/python3.9/site-packages/gi/importer.py", line 33, in <module>
    from .overrides import load_overrides
ImportError: cannot import name 'load_overrides' from 'gi.overrides' (/var/home/thiblahute/devel/gstreamer/gstreamer/subprojects/gst-editing-services/bindings/python/gi/overrides/__init__.py)
Factory Details:
```

The approach to fixing it is to implement override `gi` in
`gst-python/gi/` which we add to `PYTHONPATH`) and in there reset the
`gi` module to the right place and we get overrides from paths from
`_GI_OVERRIDES_PATH` we set in `gst-env.py` which points to all the
overrides that will be installed.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1155>

mediafoundation: Fix various string constness handling

... with fixing typo (g_strup -> g_strdup)

Constness needs to be explicit in C++ world otherwise compiler
would complain about that.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1164>

gstreamer/gst/gstconfig.h.in: Add support for ARC64 architecture

Signed-off-by: Veronika Kremneva <kremneva@synopsys.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1161>

missing transceiver unref in WebRTC renegotiation example

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1160>

Docs: Fix a typo for Intel's vaapi driver name.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1151>

wraps:ffmpeg: Move to 4.4

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1156>

tests:va: Fix null ptr dereference in multi-vpp

Dereference the pointer err before null check, which raised a null
pointer dereference issue by Coverity. Modify it to do the null check
of err first, then dereference it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1143>

va: Fix null ptr dereference for vadeocder

Making a null check in gst_va_decode_picture_free () indicates pic->buffers or pic->slices
can be null, then in _destroy_buffers () the pointers are dereferenced, which is detected
as dereference after null check by Coverity. Thus, modify the code to do null check in
_detroy_buffers ().

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1143>

Pin all wrap files to closest tag or commit sha1

Tracking branches, especially master, can cause regressions at any time.
We should try to use as much stable release tags as possible in the
future.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1152>

tsdemux: Handle delayed seek events

Store the event in case it cannot be processed immediately and process
it after the first segment has been produced.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/980>

tsdemux: Protect demux->segment_event with a mutex

Would otherwise cause weird issues when processing a delayed seek event

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/980>

video-converter: Fix v210->I420 last line conversion

Last line would not be converted correctly if height was an odd number.
Fixed by accounting for data type (8bit vs. 16bit) differences between
respective packing and unpacking functions.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/973>

video-format: Clean up v210 packing code

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/973>

video-converter: Add fast paths from I420/YV12, UYVY, Y42B and YUY2 to v210

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/973>

codecs: h264: Fix wrong type of ret variable

This ret is not a GstFlowReturn. This broke v4l2 decoder which does not
implement new_picture() virtual function.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1154>

d3d11: Rename screen capture element

Old name "desktopdup" may confuse users. Now it's renamed to
"screencapture"

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1136>

gitignore: Ignore *.wrap files again

There are various redirection *.wrap files which result in noisy
diff. Just ignore the entire wrap files. We can update wrap files
anytime via -f option.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1140>

ci: Move rules from `.build fedora x86_64` to `.build`

They should apply to all job that extend it as the 2 jobs that depend on
`.build` but not on `.build fedora x86_64` is
`build nodebug fedora x86_64` which pulls `build static nodebug fedora x86_64`
both jobs should be manual on `main` and they should run only if some
files "modules_changes" happened

Also make sure we do no automatically run job on `main` branches from forks

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1141>

ci: Fix ordering of rules to always build on main

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1141>

hooks:python: Max line length should be 120

As we ignore line length in pycodestyle and we have many places with
line with more than 80 and less than 120 chars

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1121>

validate:launcher: Reindent utils.py

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1121>

validate:launcher: Check if the running MR will close a referenced issue

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1121>

ci: Check for closed bugs for unit tests too

And add FIXMEs for test that should have been reenabled but haven't,
we will reenable them when CI is stable enough!

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1121>

tests: Move all unit tests issues to the GStreamer monorepo

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1121>

transcoder: Set state back to NULL after run() finishes

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1063>

transcoder: Use full path for includes in 'gsttranscoder.h'

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1063>

va:display: Don't close an fd with negative value

Cannot pass negative parameter to close() and thus no need to apply
close() when fd < 0.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1131>

va: Fix error handling for decoder

Need to check if va decoder is closed successfully.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1131>

va: Fix error handling for base transform

Need to check the returned value of gst_buffer_pool_set_active() when
setting the active status of buffer pool.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1131>

Fix GStreamer msys2 builds in the CI

Install ninja and the toolchain in the docker image.
Use the UCRT64 environment for building

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1000>

ci: prepare the Windows image for MSYS2 cerbero builds

 * Install MSYS2 and VS with choco
 * Install Windows SDK 8.1 requied for libvpx
 * Improve build time of the windows image using as much RUN layers as
   possible

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1000>

gl/dmabuf: add some debug logging about why things may fail to be uploaded

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1145>

playbin2/3: autoplug/caps: don't expand caps to ANY

Retrieving the pad template caps from a ghost pad returns ANY which when
merged with any other caps will return ANY.  ANY is not very specific
and may cause suboptimal code paths in e.g. decoders that assume the
lowest common denominator when presented with ANY caps.

Fixes negotiating dma-buf with vaapidecodebin between glupload in the
video sink element.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1144>

meson:avtp: Error out if sock_txtime is not present and avtp is enabled

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1142>

Revert "tests: Blacklist some tests that recently failed"

This reverts commit 9f13ee5b0eada5ca0997e59be2ef46d5c2dd1dcd as
it was merge as part of the bigger CI gardening MR
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1139>

soup-stub: Gate G_URI_FLAGS_SCHEME_NORMALIZE behind glib version check

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1138>

ci: Run tests post merge

This will make tracking new races much easier

Mark other jobs as manual when pipeline is running on master

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1129>

ges:test: Document some GES failures

Namely races in:
 * check.gst-editing-services.edit_while_seeked_with_stop_it
 * check.gst-editing-services.check_layer_activness_gaps_it

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/803
and https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/804

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1130>