[Network] Improve internal audio on/off handling

[dai-shi]

Currently, audio track is added regardless of "Mic On" flag, which allows users to speak instantly after they enable the mic.
This is done by track.enabled = true/false. ref: https://developer.mozilla.org/en-US/docs/Web/API/MediaStreamTrack

It looks like even if the track is not enabled, the RTCSender transmits some bytes. Can we improve it by using sender.replaceTrack instead? How can we make an empty audio track at the start?

[dai-shi]

new AudioContext().createMediaStreamDestination().stream;

to create an empty audio stream.

We could use Audio API to create stream/track, instead of sender.replaceTrack.