feat(enrichment): loose MusicBrainz search fallback (find aliased/romanized artists) (#771)

* feat(enrichment): loose MusicBrainz search fallback (find aliased artists)

The MB text search used a strict field-phrase query
(`recording:"<title>" AND artist:"<artist>"`). A field phrase only matches
MusicBrainz's *primary* artist/title — it never searches ALIASES — so a
recording stored under a non-Latin primary name (大橋純子) whose romanized
form ("Junko Ohashi") is only an alias returns ZERO results, even though MB
has it. Whole swaths of a community library (e.g. romanized J-pop / city-pop
charts) were unsearchable.

- `build_recording_query(..., loose=True)` drops the field scoping + phrases
  for plain AND-ed term groups (`(telephone number) AND (junko ohashi)`),
  which searches the whole document incl. aliases.
- `_mb_search_recordings` runs the strict query first (unchanged, high
  precision) and only on an EMPTY result retries once with the loose query —
  so mainstream matches are untouched and the extra throttled request is spent
  only on a miss. Results are re-scored by rank_candidates, so recall goes up
  without lowering match quality (auto-accept still needs the per-field floors).

Verified live: "Junko Ohashi / Telephone Number" and "Anri / Windy Summer"
(both 0 under the strict query) now surface the real records; "AC/DC /
Highway to Hell" still hits strict at score 1.0 with no loose retry.

Follow-up (separate): alias-aware SCORING so these can auto-confirm, not just
appear as manual candidates.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(enrichment): keep live exclusion in the loose search fallback

The loose fallback dropped the strict path's -secondarytype:Live filter, so a
studio chart whose strict query missed could fall back to — and, since
score_candidate doesn't penalize live takes, auto-confirm — a live-only
recording. Apply the same live gate to the loose query (skipped only when the
source title is itself a live take).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>

lib/mb_match.py

@@ -234,12 +234,37 @@ def _lucene_escape_phrase(s: str) -> str:
 _LIVE_GROUP_RE = re.compile(r"[(\[][^)\]]*\blive\b[^)\]]*[)\]]", re.IGNORECASE)
 
 
-def build_recording_query(artist, title) -> str:
+def build_recording_query(artist, title, *, loose: bool = False) -> str:
     """Lucene query for /ws/2/recording. Built from the DENOISED fields —
     the noise we strip (author credits, "(Live)", "(v2)") would otherwise
-    poison the search server's own scoring."""
+    poison the search server's own scoring.
+
+    ``loose=True`` drops the field-scoped quoted PHRASES for plain AND-ed
+    term groups (``(telephone number) AND (junko ohashi)``). The point:
+    a field phrase like ``artist:"Junko Ohashi"`` only matches MusicBrainz's
+    *primary* artist name — it never searches ALIASES — so a recording stored
+    under a non-Latin primary (大橋純子) whose romanized name is only an alias
+    is invisible to the strict query. A loose term query searches the whole
+    document, aliases included, and surfaces it. Lower precision by design: it
+    is a FALLBACK for when the strict query returns nothing, and its results
+    are re-scored by ``rank_candidates`` (and, for auto-match, gated by the
+    per-field floors), so noise never auto-applies."""
     t = denoise(title)
     a = denoise(artist)
+    if loose:
+        # denoise() already reduced each field to lowercase [a-z0-9 and] tokens
+        # (punctuation → spaces, diacritics stripped, & → "and"), so no
+        # Lucene-special character survives to need escaping. Group each field's
+        # terms and require both groups.
+        q = " AND ".join("(%s)" % g for g in (t, a) if g)
+        # Keep the SAME live exclusion as the strict path: the loose query is
+        # lower-precision, and score_candidate doesn't penalize a live take, so
+        # without this a studio chart whose strict query missed could fall back
+        # to — and auto-confirm — a live-only recording. Skipped only when the
+        # source title is itself a live take (mirrors the strict path).
+        if q and not _LIVE_GROUP_RE.search(str(title or "")):
+            q += " AND -secondarytype:Live"
+        return q
     parts = []
     if t:
         parts.append('recording:"%s"' % _lucene_escape_phrase(t))