chore(tide-table): regenerated tf-idf + locate at df<=40, and ignore the rebuildable caches

sample_tfidf.json re-run over the corpus; set_finder_locate.json is the 35-bank
pass (was 19) after MAX_DF rose from 6 to 40 on the strength of 1/df weighting.

The npz hash DB (2.3 MB, rebuilds in 39 s) and the per-recording query profile
are ignored: they are caches keyed on inputs, not sources. Noting the tension
deliberately — 'generated + gitignored + never compared' is exactly what made
preload staleness debut at a venue — so both are produced by an explicit
subcommand and their inputs are versioned.

.gitignore

@@ -32,3 +32,12 @@ output/Take95/*.flac
 output/Take95/tracks/
 output/Take95/album-mp3/
 output/Take95/*.log
+
+# set_finder: regenerable analysis artifacts, not sources.
+# The hash DB is 2.3 MB and rebuilds in 39 s; the query profile is a per-recording
+# cache keyed on mtime. Committing them would put binaries in the history for no
+# gain — but preload staleness taught us the cost of "generated + gitignored +
+# never compared", so both are rebuilt by an explicit command, never implicitly.
+armada/tide-table/set_finder_db.npz
+armada/tide-table/.qprofile_*.npz
+armada/tide-table/set_finder_scan.json