Bug 253997

This blocks bug 225367.

Currently, MockSourceBufferPrivate::append process init and media segments in the same loop. [1]

Immediately after parsing the init segment, it calls SourceBufferPrivate::didReceiveSample

SourceBufferPrivate::didReceiveSample will drop any frames with an unknown track buffer ID. [2]

This is fine when the MockSourceBufferPrivate runs in the content process and the entire append operation is synchronous.
However, when the operation occurs in the GPU Process, the track id will not be usable until the initialisation segment has been sent back to the content process to be processed by the SourceBuffer::sourceBufferPrivateDidReceiveInitializationSegment [3] so that it can create a new TrackBuffer.

the SourceBufferPrivateAVFObjC will queue all demuxed frames until after sourceBufferPrivateDidReceiveInitializationSegment has completed (which is an asynchronous operation). and only call SourceBufferPrivate::didReceiveSample once done [4]

[1] https://searchfox.org/wubkat/rev/23d40f46815dafd5828fa1932569943b85f8f24e/Source/WebCore/platform/mock/mediasource/MockSourceBufferPrivate.cpp#141-157
[2] https://searchfox.org/wubkat/rev/23d40f46815dafd5828fa1932569943b85f8f24e/Source/WebCore/platform/graphics/SourceBufferPrivate.cpp#670-679
[3] https://searchfox.org/wubkat/source/Source/WebCore/Modules/mediasource/SourceBuffer.cpp#856-857
[4] https://searchfox.org/wubkat/rev/23d40f46815dafd5828fa1932569943b85f8f24e/Source/WebCore/platform/graphics/avfoundation/objc/SourceBufferPrivateAVFObjC.mm#466-474