speaker identification from multitrack audio

In some scenarios the audio is available in a multitrack format (one speaker per microphone). We should have the ability to ingest this as both one audio file with multiple channels and as multiple audio channels with separate files and derive speaker identification from this.

The issue names no files, tests, or entry points, so first trace the audio-ingestion and speaker-identification paths in the repository. Done should include support for one multichannel file and multiple per-speaker files, with speaker identification derived from both formats.

Written by the indexing model from the issue text.