Verify correctness of audio-level and mid handling, since we’ve switched to upstream’s implementation of those.

We're getting audio levels in getSynchronizationSources, and I see mid/audio-level in the RTP. There is one potential wrinkle though; the mid extension is not sent for all RTP packets, only the first couple of packets (although in every case, I see a receiver report for the SSRC before the mid extension is omitted; maybe libwebrtc is waiting for RR before omitting?). I could see this being problematic if there is packet loss early on.

What do you think?

As a test, I tried dropping all outgoing RTCP, and the RTP packets all had the mid extension present. It certainly looks like libwebrtc is waiting to see an RR before omitting the mid extension, which is something at least.