va: display: Optimize out some property indirection

Because it was visible during some profiling, I thought it cost nothing
to optimize out the uneeded property get roundtrip.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8179>

avcodecmap: Add some more channel position mappings

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8173>

wpe: enable console message with WPE2

Looks like the WebKitConsoleMessage API is now available in WPE2 as well:
https://webkitgtk.org/reference/webkitgtk-web-process-extension/stable/signal.WebPage.console-message-sent.html

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8169>

va: h264dec: Allow "extended" profile decoding

Extended is identical to main but allows for FMO/ASO features to be
used, and prevent using CABAC. Using similar logic to "baseline",
assume that if we support main, we can also do extended.

This fixes the following fluster vectors, which otherwise would fail
when trying to link the parsebin pad.

  - BA3_SVA_C
  - MR6_BT_B
  - MR7_BT_B
  - MR8_BT_B

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8164>

pad: Only remove TAG events on STREAM_START if the stream-id actually changes

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4097

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8135>

webrtcbin: Tear down src and sink bins before removing them from webrtc

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7900>

webrtcbin: Fix potential deadlock on bin elements cleanup

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7900>

vtenc: Fix authors of encoder features

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8120>

vtenc: Fix class hierarchy in an attempt to fix property docs

Also fix some convention-nits in the process.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8120>

vtenc, osxaudio: Fix missing since markers

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8120>

vtenc: Mark rate-control enum as plugin API, and update cache

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8120>

atenc: Add missing since markers and mark plugin API

Also update the plugin docs cache.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8120>

avfvideosrc: Add missing since markers for screen-crop properties

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8120>

docs: Update macOS plugin docs again

Contains the following updates:

* New properties on avfvideosrc: screen-crop-*
* H265 and H265 Alpha support in vtdec and vtenc (VideoToolbox)
* ProRes support in vtenc
* New properties on vtenc elements: rate-control, data-rate-limits,
  max-frame-delay
* New plugin atenc (AudioToolbox) with support for encoding AAC
* Plugin move: atdec moved from -bad to -good
* New property on osxaudio elements: unique-id
* OS X -> macOS

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8120>

meson: Skip the plugins_doc_caches target if docs are disabled

This is a follow-up to 95eca6d91912620a8a36b55ae32f72a32bc2475b, due
to which the target existed, but was a no-op, and it was difficult to
notice why the plugin caches are not being generated, since the doc
option is disabled by default.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8120>

cccombiner: expose new input-meta-processing type, force

In force mode, generated captions are discarded even if input video
buffers do not hold CC meta.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8091>

meson: Improve NVMM CUDA detection

1. Add some comments explaining what headers and libs are expected on
   what systems
2. Only look in default incdirs if no incdir is specified
3. Require libnvbufsurface.so on Jetson when cuda-nvmm=enabled
4. Require libatomic on Jetson when cuda-nvmm=enabled

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8021>

info: Use an RW lock to protect the log functions and their list

Previously the code tried to be thread-safe by only locking when modifying the
list and leaking the old list, but this was not sufficient. When removing a log
function, its user_data would be freed but this log function and its user_data
might afterwards still be used during logging which then could lead to memory
corruption.

To avoid that, use an RW lock: get a write lock whenever modifying the list and
get a read lock whenever only using the list of log functions for logging.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3660

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7151>

d3d12: Suppress misleading leak report

Set may-be-leaked flag to child objects if needed,
since the parent object holding refcount of the child
will be leaked intentionally

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8152>

ges-launcher: return early on 'project-loaded' signal on error

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4102

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8145>

ges: project: Add missing API guards

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8145>

ges: formatter: Keep a reference on the timelines

We will be using it in an async way so we need to keep our own
reference to it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8145>

d3d12: Add d3d12deinterlace element

Adding D3D12 compute shader based deinterlace element
with YADIF filtering

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8140>

d3dshader: Add YADIF deinterlacing compute shader code

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8140>

appsrc: Decrease log level for item drop

Leaky appsrc feature was implemented similar to queue. Dropping item
is a normal case because it is configured so. Warning messages
is too severe for this event.

Level changed to DEBUG just like in queue.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8147>

ges-meta-container: Fix the GET_INTERFACE macro

It had a synthax issue

Fixes https://gitlab.freedesktop.org/gstreamer/gst-editing-services/-/issues/141

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8146>

videodecoder: set decode only flag by decode only buffer

This is used for optimize frame drop issue. Some video streams such as av1
contains decode-only frames. These frames are handled by decoder but not show
by display, which will cause input frame count unmatch the output frame count.
After collecting more decoder only frame, it will cause error report "so many
frame droped". So attach the flag GST_BUFFER_FLAG_DECODE_ONLY with the input
buffer, the v4l2videodecoder can further handle the decode-only buffer in
output list.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8142>

urisourcebin: Only rewrite stream-start event once

If the custom field is already present, we don't want to rewrite the event
again. Otherwise it would cause downstream into believing that a *new*
stream-start event has arrived ... which isn't the case.

Fixes #4097

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8132>

gl: raise WARNING instead of ERROR when no connector is connected

This is a valid and working use case when rendering on a headless
server.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8136>

rtppassthroughpay: ensure buffer is writable before mapping writable

It is entirely possible that the incoming buffer into _chain() is not writable
and will result in a critical when trying to map().

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8131>

analytics: add rotation to object detection mtd

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7938>

webrtc: Simplify fmtp handling in codec stats

Parsing the whole caps as SDP media only to retrieve the fmtp field afterwards
seems a bit superfluous. By looking up the a-fmtp attribute directly the number
of allocations in this function gets down a bit.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8125>

meson: unset GST_TRACERS for g-ir-scanner to avoid warnings

People might have GST_TRACERS=leaks set in their environment
by default, which will now trigger criticals during the build
when calling g-ir-scanner, because we unset GST_PLUGIN_SYSTEM_PATH
so that the scanner doesn't load any plugins.

Fixes #4093

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8121>

vkoperation: enable inline query only if it's a video operation

This commit enables the usage of inline queries, if and only if, the
provided
pNext structure, in gst_vulkan_opeation_enable_query(), chains a
VK_STRUCTURE_TYPE_VIDEO_PROFILE_INFO_KHR typed structure.

Also it guards "gstvkvideo-private.h" include

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8112>

vkvideo: add video_maintenance1 check

Add gst_vulkan_video_maintenance1_supported
to check if the video session needs
VK_VIDEO_SESSION_CREATE_INLINE_QUERIES_BIT_KHR

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8112>

avviddec: Renegotiate srcpad caps on framerate change

We avoid resetting the internal FFmpeg decoder on framerate changes,
but in turn this means we were not updating the framerate on the srcpad,
which was clearly incorrect. This change keeps the optimization but ensures
that we renegotiate downstream when framerate changes occur.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8118>

decklink: Add missing video modes to gst_decklink_mode_get_type()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8110>

decklink: Fix copy-paste errors regarding 8K modes

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8110>

doc: Add some missing doc strings

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

docs: Do not try to generate cuda documentation when gir is not generated

On macos it is not

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

tracers: leak: Fix some docstrings

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

doc: Add tracer objects information

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

tracers: latency: Port to using property based configuration

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

gst-inspect: Document tracer properties

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

typos: Fix avalaible/available typos

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

tracers: rusage: Mark as set_use_structure_params

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

gst-plugin-scanner: Do not create tracers when inspecting plugins

Ensuring that GST_TRACERS is not set in that process

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

leaks: Port to using object property based parameters

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

tracers: Simplify params handling using GstStructure and object properties

Instead of having each tracer implement its own parameter parsing,
centralize the handling in the tracer subsystem using GstStructure.
This simplifies tracer implementations and provides a consistent way
to handle properties.

It also allows for much better documentation by forcing tracer object
to expose properties

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8086>

uridecodebin3: Don't use decodebin3 static sink pad

Makes the code leaner, and fixes the issue which would happen when the
urisourcebin stream that was linked to the sink pad goes away (without any
replacement).

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7949>

decodebin3: Make usage of static sink pad optional

There is no reason why we should mandate people to "at least" use the static
sink pad. This caused issues, like mandating that it should always have valid
content linked to it (problematic in case of upstream stream changes).

Instead we only use it if it's actually linked to, in which case it gets added
to the list of inputs.

This actually simplifies the code too.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7949>

urisourcebin: Fix issue re-using outputs

When checking whether a no-longer used output could be re-used for another slot,
we only want to do that for streams which are not still used.

Otherwise we end up potentially re-assigning a demuxer stream to a completely
different one

Furthermore, if we *are* re-using an output slot, indicate what the replacement
GstStream will be so slot matching can work properly (which can happen in the
case of demuxers which add/remove all pads even if only a single one changed)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7949>

analytics: update gir

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8087>

analytics: add _N_TO_N relation type

This relation type define relations between each components of two groups.
First component of first group relate to first component of second group,
Second component of second group relate to second component of second group,
and so on. It's a denser way to express relations in this context.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8087>

vecdeque: Fix documentation of push_tail_struct()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8111>

rtpsession: Fix twcc stats structure leaks

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8073>

curlhttpsink: Set auth any for http_proxy and https_proxy

There was different behaviour if the proxy was configured through
properties or environment. For properties libcurl would be configured
with any auth, but for environment libcurl would default to using basic.
Now any auth is set for both configuration methods.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7935>

curlhttpsink: Don't set Content-Length to 0 for proxy

The Content-Length header would unconditionally be included when the
proxy property was set. This would result in requests with both
Content-Length and Transfer-Encoding header. Now we rely on the
use-content-length property in the proxy case aswell. This also makes
sure that Content-Type is set correctly, since before that would be
skipped if proxy was used.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7935>

onnx: disable onnxruntime telemetry

Disable the telemetry feautre on onnxruntime.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8092>

v4l2object: move height setting outside of per-plane loop

In the V4L2 format, the height parameter
is an attribute of the entire image, not of an individual plane.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8088>

gstvalue: fix crash in transform allocation params to string

when gst_buffer_pool_config_set_allocator (config, alloc, NULL);
gst_structure_to_string or GST_DEBUG (pool, "config %" GST_PTR_FORMAT,
config) will crash. this patch fix that

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7943>

splitmuxsrc: Calculate the end offset of a fragment once known

If a new fragment is added with a valid duration but no offset then the start
offset is set later based on the end offset of the previous fragment. At that
point the end offset of this fragment can also be calculated and not doing so
would give the next fragment the same start offset.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8069>

build: gl: Properly add sub-package to gst_libraries

GstGLEGL/Wayland/X11/Viv was actually never written into any of the gir_targets,
as a side effect, their gir was never copied into girs/ and not used to generate
the documentation.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8061>

zxing: Update decode hints usage for compatibility with ZXing >= 2.2

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7879>

flacparse: Improve header validity checks

Allow sample rate, number of channels and bps to change and in that case update
the caps accordingly.

Also move (non-fatal) validity checks and storing of the header values outside
the actual parsing once we actually know that a valid frame is available.

And also don't warn on the last frame with fixed block size blocking strategy
that the block size has changed: the last frame is allowed to be smaller.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3281

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8075>

h265parse: reset nalparser to NULL after it is freed

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8074>

h264parse: set nalparser to NULL after it is freed

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8074>

h265parse: remove useless NULL setting in gst_h265_parser_free

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8082>

h264parse: remove useless NULL setting in gst_h264_nal_parser_free

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8082>

video: fix AV12 format lacking the GST_VIDEO_FORMAT_FLAG_ALPHA flag

The GST_VIDEO_FORMAT_FLAG_ALPHA indicates whether a video format has an
alpha channel.  AV12 does have an alpha plane.  Switching from
MAKE_YUV_FORMAT to MAKE_YUVA_FORMAT to generate GST_VIDEO_FORMAT_AV12.

Fixes #4081

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8067>

urisourcebin: Fix collection leak

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8080>

tsparse: Extract and fill m2ts header for each packet

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7625>

ci/abi-check: update image for vulkan changes

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7995>

vkoperation: use inline query with video maintenance1

When video_maintenance1 is supported,
gst_vulkan_operation_begin_query will now use
the inline query mechanism instead of vkCmdBeginQuery
API.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7995>

vkdevice: enable VK_KHR_VIDEO_MAINTENANCE_1_EXTENSION_NAME

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7995>

vkphysicaldevice: dump if video maintenance1 is supported

Dump if VK_KHR_video_maintenance1 features is supported by the driver.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7995>

libav: free protocol info in gst_ffmpeg_pipe_close

When the media is not seekable, we are calling gst_ffmpeg_pipe_close
which does free the info.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8070>

v4l2codecs: Use GstVideoInfoDmaDrm more consistently

This avoids some duplications and makes the DRM info available in
more places, which will help with future changes.

Also fix some error messages while on it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8028>

subprojects: ntv2: Add check for libudev

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4079

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8068>

h264parse: Free SEI if parsing succeeds but alignment afterwards fails

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8056>

h265parse: Free SEI if parsing succeeds but alignment afterwards fails

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4076

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8056>

meson: Drop max version bound from neon

Neon 0.34.0 broke the build again, but the API+ABI has been stable since
0.27 and the library is so-versioned.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8008>

h265parse: parse unregistered SEI without user data

Same change as in h264parse.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7931>

h264parse: parse unregistered SEI without user data

We get loads of warnings when parsing videos from users:

gsth264parser.c:1115:gst_h264_parser_parse_user_data_unregistered: No more remaining payload data to store
gsth264parse.c:646:gst_h264_parse_process_sei:<h264parse0> failed to parse one or more SEI message

Those are raised because of unregistered SEI without user data.

The spec does not explicitly state that unregistered SEI needs to have
data and I suppose the UUID by itself can carry valuable information.
FFmpeg also parses and exposes such SEI so there is no reason for us no
too as well.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7931>

examples: iOS: Remove frei0r plugin as we do not ship its plugins

See https://gitlab.freedesktop.org/gstreamer/cerbero/-/merge_requests/1651#note_2681739

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8024>

tsdemux: Lower a GST_FIXME to a GST_DEBUG

This is not really a refcounting issue and can happen if a new program is in the
process of being activated that contains streams with the same PIDs.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7898>

ci/abi-check: update image for vulkan encoder changes

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: add gst_vulkan_encoder_is_started()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: add again GST_TYPE_VULKAN_ENCODER_RATE_CONTROL_MODE

It was already part of the old rate control mechanism but it had wrong the
namespace.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: implement callback to chain control rate structures

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: implement callback to chain codec specific structures

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: handle quality level

It creates a new structure for passing the codec quality structure at _start(),
where it will be filled. The quality level can be set or changed according
encoder limits.

Later the quality level will be set at _update_session_parameters() and at each
frame encoding. That's why it has to be set at _start().

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: rename first_encode_cmd to session_reset

Since it reflect better when it's needed to be used: to reset the current
session.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: shuffle down VkVideoBeginCodingInfoKHR initialization

to make it more cohesive

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: remove rate control handling

It will be reintroduced later with different approach.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: enhance algorithm to get the slot index

The algorithm for generating the current slot index is a simple round robin,
nonetheless it's not assured that the next slot index it's not still used by a
still living encode picture.

This new way holds an array with the still living encode pictures and the next
slot index looks for a released index in the array.

Its downside is deallocating a picture need to be removed from the array, so the
helper has to be passed to the uninit() function

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: add VkVideoReferenceSlotInfoKHR in GstVulkanEncoderPicture

And remove slotIndex since it's part of VkVideoReferenceSlotInfoKHR, simplifying
the reference slots array creation, and changing the tests accordingly.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: check for codec buffer

gst_vulkan_video_codec_buffer_new() can return NULL, so it's required to check
the returned value and bail out if needed.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: avoid GstVulkanEncoderPicture allocation

By using it as apart of the encoder picture structure that has to initialized
and uninitalized.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>

vkencoder-private: remove width, height and fps from GstVulkanEncoderPicture

In GStreamer that buffer information is decoupled, holding other structures to
describe the stream: GstCaps. So, to keep the GStreamer design this patch
removes these information from GstVulkanEncoderPicture and pass to
gst_vulkan_encoder_encode() a pointer to GstVideoInfo.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8007>