[FEAT]: ds.extract_stems(audio_path, drumless=True, verbose=True) creates a drumless backing track but does not return its path. #298

Summary

ds.extract_stems(audio_path, drumless=True, verbose=True) creates a drumless
backing track but does not return its path.

Detail

extract_stems() (drumscript/init.py) calls separate_audio(), which
returns a dict containing the backing-track path under the key mix. But
extract_stems() discards that dict and constructs its own:

{"status", "drum_stem_path", "original_file", "output_directory"}

so mix never reaches the caller. Verified at runtime — the returned keys are:

['drum_stem_path', 'original_file', 'output_directory', 'status']

The backing track exists on disk (as <input>_no_drums.<ext> inside
output_directory) but the user has to reconstruct that path by hand.

Fix

Carry the path through from separate_audio()'s return value, exposed under a
readable key backing_track_path (not the terse mix). Then update the README
backing-track example to print remove_drums['backing_track_path'].

Target

v0.2.1

Notes

• v0.2.0 ships a doc-only workaround: the README example uses output_directory.
• Same class of bug as the build_score() / transcribe() path-reporting fix
  in v0.2.0 — a function doing more than it reports.