It occurred to me to wonder whether it was really Description, Source,
or both that results in skipping the map_vbs_exception logic.

It turns out we had it right (it's just description, and it's
null-pointer and not empty string that makes the difference).

But the fact that it's not obvious still makes it a good testcase

Add test for E_UNEXPECTED.

Based on patch by Kevin Puetz.

Fixes a regression introduced by commit
6aa437de.

This can happen with asynchronous decoding, such as with the VA-API
decodebin plugin for H264.

Allowing easier preview of NV12 / I420 images as well.

Keeping them in a IMFCollection.

Since struct _IAVIStreamImpl has a pointer to a WAVEFORMATEX,
GCC 12.2 emits warning when dereferencing that pointer when
the block has been allocated with sizeof(PCMWAVEFORMAT).

The warning is fixed by always allocating with sizeof(WAVEFORMATEX).
This will overallocate in case of a PCM stream.

The alternative would have been to store in struct _IAVIStreamImpl
a pointer to PCMWAVEFORMAT instead, and add the casting to a
WAVEFORMATEX when needed. That would clutter the code IMO since most
of the ACM APIs expect a LPWAVEFORMATEX.

/home/eric/work/wine/dlls/avifil32/acmstream.c: In function 'AVIFILE_OpenCompressor':
/home/eric/work/wine/dlls/avifil32/acmstream.c:105:24: warning: array subscript 'struct tWAVEFORMATEX[0]' is partly outside array bounds of 'unsigned char[16]' [-Warray-bounds]
  105 |       This->lpOutFormat->wFormatTag = WAVE_FORMAT_PCM;
      |                        ^~
/home/eric/work/wine/dlls/avifil32/acmstream.c:101:27: note: object of size 16 allocated by 'HeapAlloc'
  101 |       This->lpOutFormat = HeapAlloc(GetProcessHeap(), 0, This->cbOutFormat);
      |                           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Signed-off-by: Eric Pouech <eric.pouech@gmail.com>