Suspend PulseAudio sinks when they are not used

[savchenko]

Describe the bug
Mumble keeps PA sink awake even when client has its microphone and speakers disabled ("deafened and muted").

Steps to Reproduce

1. Launch pulseaudio -v and observe its output.

2. Compare, for example, with mpv or your favourite music player. Start playing, then put it on pause:

   I: [pulseaudio] module-suspend-on-idle.c: Sink alsa_output.usb-Solid_State_Logic_SSL_2-00.analog-stereo idle for too long, suspending ...
   I: [alsa-sink-USB Audio] alsa-sink.c: Device suspended...
   I: [pulseaudio] core.c: All sinks and sources are suspended, vacuuming memory
   

3. Close the audio player, try the same with Mumble.

4. Sink stays open:

   I: [pulseaudio] protocol-native.c: Requested tlength=60.00 ms, minreq=10.00 ms
   I: [pulseaudio] protocol-native.c: Final latency 60.00 ms = 20.00 ms + 2*10.00 ms + 20.00 ms
   

   This results in an excessive CPU usage and unnecessary memory reservation. Former is (much) more of a problem as for some reason running Mumble makes PA eat ~10% CPU. For comparison, playing audio via mpv is about 0.5-1%.

Expected behavior

1. Mumble tells PA to pause the processing when it's not needed.
2. Ideally, PA/Mumble combination consumes (much) less CPU than 10% of 11th gen i5 :)

Screenshots
N/A

Desktop (please complete the following information):

• OS: Debian 11, pulseaudio 14.2.
• Version: master with FIX(client): Disable broken shortcuts on Wayland #5307 applied.

Additional context
If #5300 is fixed, this might be irrelevant (or at least much less relevant). Modern ALSA does all the mixing and resampling by itself.

[Krzmbrzl]

Related to the high CPU usage: #4177 and #1092

In terms of this request itself: I think this could be a (partial) duplicate of #1089 which was implemented for PulseAudio in #171 but reverted again in #4633, because this has caused quite a few issues due to the way the audio system is currently implemented.

The gist of it is: The audio system needs to be refactored and redesigned before something like this is possible.

Given that we already have #1089 (and the other linked issues), I think this issue here does not really describe a separate issue. Therefore, I'll close this in favor of the other issues.