Update to webrtc_track_id function for handling case without msid-semantic

[dguerizec]

This pull request updates the webrtc_track_id function in the SessionDescription class, found within sdp.py.

The main change implemented is an additional check for the scenario where msid-semantic is not provided. The updated code adds a check for this scenario, and in such cases, it directly returns bits[1] as the track id.

Prior to this update, the function would return None in scenarios where msid-semantic is not present in OFFER, as it strictly expected at least a msid-semantic to return a track id when an msid was present. This change thus enhances the function's flexibility and robustness in handling cases where media streams have an msid but no msid-semantic field is present.

This change was needed on my side to be able to connect an aiortc peer to Livekit SFU: Livekit does provide an msid for each track, but does not provide the msid-semantic field.

After some research, I found that it seems that the msid-semantic is not required (and may have been obsoleted between draft 08 and 09 of RFC-8830, see references below).

Here's a breakdown of the changes:

Added a condition to check if self.msid_semantic is non-empty.
If self.msid_semantic is empty, bits[1] (track id) is directly returned.
If self.msid_semantic is non-empty, the existing behavior is retained.

This update ensures that the webrtc_track_id function can handle all scenarios more effectively, returning a valid track id even in cases where msid-semantic is not present.

References:

• msid-semantic description in draft version 08:
  RFC-8830 (draft#08)

• msid-semantic attribute removed in draft version 09:
  RFC-8830 (draft#08)

[dguerizec]

I also pushed some new fixes to make it work with Livekit SFU:

• Ensure the tracks are closed when they go in "inactive" state
• Recycle receiver with ssrc update
• Update msid on track if it changed for this mid

[jlaine]

Could we please have some unit tests, which might also help to clarify the usecase?