PSA: Refactoring statistics code of the audio processing module.

[Ivo Creusen]

If you get audio processing stats from the GetStats() interface on the PeerConnection API (like most users) there will be no changes at all, and you can stop reading.

In about 2 weeks time, I will make several changes to how audio processing statistics are passed around inside WebRTC, which involves some small changes to some of the public C++ APIs: 

- In modules/audio_processing/include/audio_processing.h, the AudioProcessingStatistics struct will be removed, and the accompanying GetStatistics() function that returns it will also be removed. These are replaced by a new struct called AudioProcessingStats (which can be found in modules/audio_processing/include/audio_processing_statistics.h), and a GetStats function that returns it.

- In api/mediastreaminterface.h, the AudioProcessorInterface::AudioProcessorStats struct will be removed, and the accompanying GetStats() function that fills it will be removed as well. These are replaced by a new struct called AudioProcessorInterface::AudioprocessorStatistics, which can be found in the same file, and a new GetStats function that can return it.

- In media/base/mediachannel.h, several members will be removed from the VoiceSenderInfo struct: aec_quality_min, echo_delay_median_ms, echo_delay_std_ms, echo_return_loss, echo_return_loss_enhancement, residual_echo_likelihood and residual_echo_likelihood_recent_max. These same values (except aec_quality_min, since it was not in use) can now be found inside the apm_statistics member in the same struct.