sequence: tracks without track_class row are excluded from pool even when enrichment exists

[gnujoow-bot]

Summary

tracks_for_sequence() JOINs track_class:

FROM tracks t
JOIN track_class tc ON tc.track_id = t.spotify_id
LEFT JOIN enrichment e ON e.track_id = t.spotify_id
...
WHERE t.is_liked = 1

If a track is inserted directly into tracks (e.g. a playlist-only track added manually) but classify has not been run, there is no track_class row, so the track is excluded from the sequence pool entirely — even when it already has full enrichment (BPM/camelot/energy) data.

Impact

resequence on such a playlist sequences only a handful of tracks and dumps the rest as "no mixing data" / original order, despite the data being present. Requires a separate classify step that is easy to forget.

Expected

Either:

• tracks_for_sequence should LEFT JOIN track_class (and treat missing buckets as permissive), or
• the enrichment/import flow should always run classify (or create a default bucket) so pool membership does not silently depend on a separate manual step.

Reproduction

1. Insert playlist tracks into tracks with is_liked=1 and enrich them (enrichment rows present).
2. Skip classify.
3. Run resequence "<playlist-url>" --create "reseq" → only 1–3 tracks sequenced.
4. Run classify, then resequence again → 148/157 tracks sequenced.