Miniaudio

=====================
* Add support for setting the device notification callback when initializing an engine object.
* Add support for more than 2 outputs to splitter nodes.
* Fix a crash when initializing a channel converter.
* Fix a channel mapping error where weights are calculated incorrectly.
* Fix an unaligned access error.
* Fix logging with the C++ build.
* Fix some undefined behavior errors, including some memset()'s to null pointers of 0 bytes.
* Fix logging of device info for loopback devices.
* WASAPI: Fix an error where 32-bit formats are not properly detected.
* WASAPI: Fix a bug where the device is not drained when stopped.
* WASAPI: Fix an issue with loopback mode that results in waiting indefinitely and the callback never getting fired.
* WASAPI: Add support for the Avrt API to specify the audio thread's latency sensitivity requirements. Use the `deviceConfig.wasapi.usage` configuration option.
* PulseAudio: Pass the requested sample rate, if set, to PulseAudio so that it uses the requested sample rate internally rather than always using miniaudio's resampler.
* PulseAudio: Fix a rare null pointer dereference.
* ALSA: Fix a potential crash on older versions of Linux.
* Core Audio: Fix a very unlikely memory leak.
* Core Audio: Update a deprecated symbol.
* AAudio: Fix an error where the wrong tokens are being used for usage, content types and input preset hints.
* WebAudio: Do some cleanup of the internal global JavaScript object when the last context has been uninitialized.
* Win32: Fix an error when the channel mask reported by Windows is all zero.
* Various documentation fixes.
* Bring dr_wav, dr_flac and dr_mp3 up-to-date with latest versions.

====================
* Fix some bugs where looping doesn't work with the resource manager.
* Fix a crash when seeking a sound.
* Fix a subtle bug the results in a glitch when looping a decoder when resampling is being applied.
* Fix an issue where chaining streams would not result in a seamless transition.
* Add a new flag called MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_UNKNOWN_LENGTH for use with resource managed data sources. This flag is used as a hint to the resource manager that the length of the data source is unknown and calling ma_data_source_get_length_in_pcm_frames() should be avoided.
* Add support for resetting a resampler. This is useful for resetting the internal timer and clearing the internal cache for when you want to seek the input sound source back to the start.
* Add support for clearing the cache from biquads and low-pass filters.

====================
* PulseAudio: Work around bugs in PipeWire:
- PipeWire is returning AUX channels for stereo streams instead of FL/FR. This workaround forces FL/FR for stereo streams.
- PipeWire will glitch when the buffer size is too small, but still well within reasonable limits. To work around this bug, the default buffer size on PulseAudio backends is now 25ms. You can override this in the device config. This bug does not exist with regular PulseAudio, but the new default buffer size will still apply because I'm not aware of a good way to detect if PipeWire is being used. If anybody has advice on how to detect this, I'm happy to listen!
* DirectSound: Increase the minimum period size from 20ms to 30ms.
* Return `MA_SUCCESS` from `ma_device_start()` and `ma_device_stopped()` if the device is already started or stopped respectively.
* Fix an incorrect assertion in the data converter.
* Fix a compilation error with ARM builds.

====================
* Fix an error when seeking to the end of a WAV file.
* Fix a memory leak with low-pass, high-pass and band-pass filters.
* Fix some bugs in the FLAC decoder.
* Fix a -Wundef warning

====================
* WASAPI: Fix a bug where the device is not stopped when an error occurrs when writing to a playback device.
* PulseAudio: Fix a rare crash due to a division by zero.
* The node graph can now be used as a node. This allows node graphs to be connected to other node graphs.
* Fix a crash with high-pass and band-pass filters.
* Fix an audio glitch when mixing engine nodes (ma_sound and ma_sound_group).
* Add some new helper APIs for cursor and length retrieval:
- ma_data_source_get_cursor_in_seconds()
- ma_data_source_get_length_in_seconds()
- ma_sound_get_cursor_in_seconds()
- ma_sound_get_length_in_seconds()

====================
* WASAPI: Fix a bug in duplex mode when the capture and playback devices have different native sample rates.
* AAudio: Add support for automatic stream routing.
* iOS: The interruption_began notification now automatically calls `ma_device_stop()`. This allows `ma_device_start()` to work as expected when called from interruption_ended.
* iOS: Fix a bug that results in a deadlock when stopping the device in response to the interruption_begain or interruption_ended notifications.
* Fix a bug with fixed sized callbacks that results in glitches in duplex mode.
* Fix a bug that results in a deadlock when starting a device.
* ma_engine_play_sound_ex() is now publicly visible.
* Add validation to ma_sound_set_pitch() to prevent negative pitches.
* Add validation to resamplers to prevent negative ratios.


---------------------------------------------------------------------------------------------------