mesa: Add MESA_SHADER_CAPTURE_PATH for writing .shader_test files.

mesa: Add MESA_SHADER_CAPTURE_PATH for writing .shader_test files.

This writes linked shader programs to .shader_test files to
$MESA_SHADER_CAPTURE_PATH in the format used by shader-db
(http://cgit.freedesktop.org/mesa/shader-db).

It supports both GLSL shaders and ARB programs.  All stages that
are linked together are written in a single .shader_test file.

This eliminates the need for shader-db's split-to-files.py, as Mesa
produces the desired format directly.  It's much more reliable than
parsing stdout/stderr, as those may contain extraneous messages, or
simply be closed by the application and unavailable.

We have many similar features already, but this is a bit different:
- MESA_GLSL=dump writes to stdout, not files.
- MESA_GLSL=log writes each stage to separate files (rather than
  all linked shaders in one file), at draw time (not link time),
  with uniform data and state flag info.
- Tapani's shader replacement mechanism (MESA_SHADER_DUMP_PATH and
  MESA_SHADER_READ_PATH) also uses separate files per shader stage,
  but allows reading in files to replace an app's shader code.

v2:  Dump ARB programs too, not just GLSL.
v3:  Don't dump bogus 0.shader_test file.
v4:  Add "GL_ARB_separate_shader_objects" to the [require] block.
v5:  Print "GLSL 4.00" instead of "GLSL 4.0" in the [require] block.
v6:  Don't hardcode /tmp/mesa.
v7:  Fix memoization of getenv().
v8:  Also print "SSO ENABLED" (suggested by Timothy).
v9:  Also handle ES shaders (suggested by Ilia).
v10: Guard against MESA_SHADER_CAPTURE_PATH being too long; add
     _mesa_warning calls on error handling (suggested by Ben).
v11: Fix crash when variable is unset introduced in v10.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>