Support temporal scalability

[lherman-cs]

L1T2 or L1T3 on each simulcast layer will allow even finer adjustments: 36% -> 25% bitrate step while allowing us to skip keyframes.

We need either frame marking or a dependency descriptor.

• a=extmap:13 https://aomediacodec.github.io/av1-rtp-spec/#dependency-descriptor-rtp-header-extension
• a=extmap:15 http://tools.ietf.org/html/draft-ietf-avtext-framemarking-07

Chrome only offers a dependency descriptor by default.

References:

• https://www.w3.org/TR/webrtc-svc/
• https://www.hhi.fraunhofer.de/en/departments/vca/research-groups/video-coding-technologies/research-topics/past-research-topics/scalable-video-coding-in-h264-avc/temporal-scalability-in-h264-avc.html