iOS change method of bypassing voice processing

[snystedt]

The implementation of AudioIODevice::setAudioPreprocessingEnabled in juce_ios_Audio.cpp uses the AVAudioSession::setMode method to set the AVAudioSessionMode of the session to AVAudioSessionModeMeasurement. However, this also removes the processing of output signals, causing a drastic reduction in output playback level (see https://developer.apple.com/documentation/avfaudio/avaudiosessionmodemeasurement?language=objc). While this might be desirable behaviour for specific use cases, the setAudioPreprocessingEnabled description in juce_AudioIODevice.h states that

/** On devices which support it, this allows automatic gain control or other
    mic processing to be disabled.
    If the device doesn't support this operation, it'll return false.
*/

suggesting the intention is to only bypass audio input preprocessing.

My proposal is to instead use the AudioUnitProperty kAUVoiceIOProperty_BypassVoiceProcessing which only disables the input processing, not the output processing. (https://developer.apple.com/documentation/audiotoolbox/1534007-voice-processing_i_o_audio_unit_proper/kauvoiceioproperty_bypassvoiceprocessing?language=objc)

If this is not desirable, then I would recommend that another method be added, AudioIODevice::setInputProcessingEnabled, which would do the same thing suggested above.

This is crucial to record singing/vocals on iOS, since the Automatic Gain Control in the phones introduce a ton of noise when not actively singing/talking, as in when you're waiting for the place you want to come in with your singing.

[snystedt]

Alright, so this was kind of a fast PR I made when discovering the issues with AVAudioSessionModeMeasurement. I'll come back to this and clean up the code. However it would be good to know if this is something that you'd consider merging in near-time. Currently I have to hack a modified juce_ios_audio.cpp into our build process and would obviously prefer to stop doing so asap. :)