feat: add drum transcription and separation f

prepare for publishing to PyPI

add audio_path backward compatibility

revert to use librosa instead of torchaudio for now