Classify the titles in und.toml by language and kind #750

Part of #747. Phase 1 converted the titles mechanically into und.toml under the generic title kind (801 words on master d3730550; recompute with uv run python tools/vocab/query.py --kind title --count).

The work: move each title into its language files and give it its kind: salutation, style or dignity, academic or professional, religious, nobility, military, office, occupation, kinship, household. Gender only where the word gives a definitive clue (decision 9).

Also: the per-word counts and reasons in decisions/salutation-titles and the other title entries move into the entries, and the prose becomes query pointers (one home per fact).

Snapshot-neutral, and independent of the particle classification, so the two can run at the same time.

Done when und.toml holds no title, the compiled sets are unchanged and the snapshot does not move.