Add mic-only muting and playback volume control to MixingAudioDeviceModule

[paterkleomenis]

⚠️ Depends on #28 — must be merged first

Summary

Adds two runtime controls to MixingAudioDeviceModule / MixingAudioControl:

• Mic-only mute — silences the microphone at the ADM level while keeping the audio channel open, so system-audio loopback continues to flow uninterrupted.
• Playback volume — scales incoming playback samples in software, providing per-call volume control independent of the system mixer.

Changes

All changes are in webrtc/webrtc_create_adm.cpp and webrtc/webrtc_create_adm.h.

MixingAudioTransport

RecordedDataIsAvailable — when _microphoneMuted is set, the mic buffer is zeroed before the loopback mix step and before forwarding to the inner transport. Loopback audio (if enabled) is mixed into the zeroed buffer, so system audio still reaches the remote peer.

NeedMorePlayData — after the inner transport fills the playback buffer, every sample is multiplied by _playbackVolume (float, 0.0–1.0). At unity gain (1.0) the branch is skipped entirely — no performance cost when unused.

Both atomics are written relaxed; the hot-path reads are relaxed too (no synchronisation needed — stale values are safe).

MixingAudioDeviceModule

New setMicrophoneMuted(bool) and setPlaybackVolume(float) methods forwarded to the active MixingAudioTransport. State is also re-applied when RegisterAudioCallback() installs a fresh transport (e.g. after an ADM restart).

MixingAudioControl (public API)

void setMicrophoneMuted(bool muted);
bool microphoneMuted() const;

void setPlaybackVolume(float volume);  // 0.0 = silent, 1.0 = unity gain
float playbackVolume() const;

State is persisted in MixingAudioControl and re-applied whenever a new MixingAudioDeviceModule attaches (via attach()), so callers can set these before the ADM is created.

Platform

Linux and Windows. macOS is unaffected.

[ilya-fedin]

The PR has conflicts

[ilya-fedin]

This PR also has the commit from #26, it's a problem

[ilya-fedin]

This PR also has the commit from #26, it's a problem

This isn't fixed

[ilya-fedin]

You made it worse. What I meant is that the commit either should go to the other PR and this PR get closed or this PR should explicitly mark that it depends on the other one. But you squashed the commits and the problem is still isn't fixed: once the other PR is merged, this one will conflict.

[ilya-fedin]

I don't see any tdesktop PR being dependent on #26. If the only reason of #26 is mic-only muting and playback volume control then just leave only one PR and name it accordingly.

[ilya-fedin]

I now see there's #27 has the same commits as #26 and this PR too...