[WebRTC] Queue frames till Video Frame Observer is ready

[ganeshprasads]

https://bugs.webkit.org/show_bug.cgi?id=317111
Reviewed by: E. Ocaña González

Problem: WebRTC playback doesnot recover from a black screen and the issue is seen Intermittently

 Cause: The first video frame(s) (containing SPS/PPS/IDR) being dropped
 because libwebrtc starts delivering decoded frames before the downstream
 GStreamer pipeline's InternalSource registers as a VideoFrameObserver.

 Without first having the SPS/PPS and the first sync frame, the decoder
 cannot decode any subsequent delta frames.

 In Following code snippet from mediastream/gstreamer/GStreamerMediaStreamSource.cpp,
 frames received while m_isObserving is false might be dropped.

void videoFrameAvailable(VideoFrame& videoFrame, VideoFrameTimeMetadata) final
{
    if (!m_parent || !m_isObserving)
       return;

    updateFirstVideoSampleSeenFlag();

Change: While this patch does not address the root cause of delayed start of
 VideoFrameObserver, add a small frame buffer to hold frames
 (currently max size set to 30 frames) that arrive before any observer registers.
 deliver the frames when observer is registered
Note: From tests it is observed that only the first frame or 2 is actually needed to be queued.

* Source/WebCore/platform/mediastream/RealtimeMediaSource.cpp: (RealtimeMediaSource::videoFrameAvailable): Queue frames till observer are ready
* Source/WebCore/platform/mediastream/RealtimeMediaSource.h: (struct PendingVideoFrame): Added a buffer to hold the frames<!--EWS-Status-Bubble-Start-->

617daa5

Build-Tests	Layout-Tests
❌ 🛠 wpe-246-amd64-build	❌ 🧪 wpe-246-amd64-layout
❌ 🛠 wpe-246-arm32-build	❌ 🧪 wpe-246-arm32-layout

[eocanha]

Ported PR #1689 (from wpe-2.38) upstream, with some modifications to avoid 4 layour test regressions, and submitted it as WebKit/WebKit#67460 for review there (on the https://bugs.webkit.org/show_bug.cgi?id=317111 upstream bug created by Ganesh).

When landed there, will backport it to 2.46 as well instead of this PR #1690.