Use `songrec listen --audio-device` instead of manually capturing audio with parec, converting with ffmpeg, and passing to audio-file-to-recognized-song. Drops temp file handling and simplifies the recognition loop.
Refactor source type handling and validation checks.