Takes audio (mp3) and a "source-of-truth" audio transcript (string) as input and returns precise timestamps.
