[jasperblues]

Closes #439. Port of liberation-data/drivine4j#622 (drivine4j PR #632), as liberation-data/drivine#647 ports it. Not to merge until train 2026.11.2 is published, and not before drivine#647.

_term_of (drivine/backend/sql/text_query.py) read a quoted phrase ending in *, "of sou*", as a prefix phrase, so it matched "of soup". Lucene's classic parser, as Neo4j uses it, reads a * inside quotes as text, so Neo4j searches for the phrase "of sou". A term is now a prefix only when its token isn't quoted. Oracle Text follows: "graph pre*" becomes ({graph pre}).

Goldens: unchanged.

Tests

• tests/integration/test_bm25.py, and its sync copy, on Neo4j, Postgres, Postgres indexed, SQLite and Oracle: "of sou*" finds nothing, and "of soup*" finds "a graph of soup".
• tests/unit/test_text_query.py:
  • the scored patterns of "of sou*", body:"of sou*", "of sou\*", "kiw*", sou* and +kiw*;
  • the Oracle Text forms of "graph pre*", "kiw*" and body:"of sou*".

Verification

• Two fresh-reviewer passes. Pass 1 added the Oracle Text cases, the class docstring and the docs/02 and docs/07 notes; pass 2 came back clean.
• Unit tests: 2,478 passed. ruff and pyright are clean, and the sync code is current.
• Integration, on the test box with Oracle: the BM25 tests, their sync copy, and the search tests. 191 passed, and agreement with Neo4j is unchanged at 17/21.