Feature request: add text to docs about Opus, not just Wav

[JRMeyer]

right now, the docs only reference training on WAV files, but we can also train on opus.

similarly, we should probably rename the expected column headers in the csv data files to something other than wav_filename,wav_filesize,transcript