Unify input/output API for RECVONLY and SENDONLY modes

[whitphx]

webrtc_streamer has many different I/O types which makes its API complex. We should get rid of some APIs so that both the API and its implementation get simpler.

Input

Currently webrtc_streamer supports 2 input types:

• Player object specified with player_factory
• Source track specified with ctx.source_video_track and ctx.source_audio_track

And #731 may also be here.

The source_*_track should be the unified input API, and the player-based API should be removed from webrtc_streamer(). Instead, the developers will manually connect the player object to the input tracks.

Output

Currently webrtc_streamer supports 3 output types:

• Recorder API with in_recorder_factory and out_recorder_factory
• Receiver API with ctx.video_receiver and ctx.audio_receiver
• Output tracks with ctx.output_video_track and ctx.output_audio_track

At least, like the player API above, the recorder API should be removed from webrtc_streamer() so that the developers manually connect the output tracks to the recorder object.

The receiver objects are kind of special, but they may also be unified with the output track-based API.

[whitphx]

However, with the current API, the players and the recorders are automatically stopped or closed by the webrtc_streamer() component. Removing these APIs also gets rid of this advantage.

-> Keep them and also add the track-based API for verbose usage? --> Only the recorder can automatically be stopped by the "stop"event emitted from the upstream track?