Web-audio-api

- Added the AudioWorkletNode and AudioWorkletProcessor interfaces
- Added stereo-to-stereo panning for PannerNode in EqualPower mode
- Aggressively release resources of 'orphaned' audio nodes
- Tuned the volume of stereo-to-stereo panning for PannerNode in HRTF mode
- Added `context.run_diagnostics` for debugging purposes
- Fixed a panic that could occur during audio graph booting (flawed `is_active` check)
- Fix: `disconnect` would disassociate an AudioNode from its AudioParams
- Fix: properly clamp PannerNode rollOffFactor according to the distance model

- Fix compilation on targets that are not x86/aarch64
- Added the ability to run benchmarks locally with `criterion`

- Disable floating point denormals in audio processing via processor flags
- Do not spawn a garbage collector thread for an OfflineAudioContext
- AudioParam now derives `Clone`

- Fix: panic when creating and dropping PannerNodes
- Improved performance of setting panning mode to HRTF

- Breaking: many AudioNode setter methods now require `&mut self` instead of `&self`
- Fix: all audio node settings are now applied in order in the audio graph
- Fix: render thread would crash when a PannerNode is removed
- Fixed and improved device ids for audio input/output devices
- Added current playhead position for AudioBufferSourceNode

- Fix: clamp to number of output channels to 32 even if the hardware supports more to prevent runtime panics
- Fix: prevent a render thread panic if the audio graph is not fully initialized yet
- Fix: rename AudioDestination max_channels_count to max_channel_count
- Fix: AudioBufferSourceNode is now Send + Sync
- Change AudioProcessor::onmessage signature to prevent deallocations
- Add garbage collector thread as sidecar of the render thread to handle some deallocations