SQL: resolve the logical name of a multi-label Cypher VECTOR/FULLTEXT index #11014

Follow-up of #10914 (PR #10994).

CREATE FULLTEXT INDEX docs FOR (n:Article|Post) ... and CREATE VECTOR INDEX over several labels or relationship types build one index per type and tie them together with a logical name that only openCypher resolves (CypherMultiTypeIndexes, a cypher.multiindex.<name> custom value on every member type).

What SQL sees today

• The per-type indexes are ordinary indexes: SEARCH_INDEX('Post[title,body]', ...) and vector.neighbors('Post[e]', ...) find the data, and a record written through SQL is found by the logical name from Cypher (pinned by Issue10914MultiLabelVectorAndFulltextIndexTest).
• The logical name (docs, emb) is not known to SQL, so a query across the types is one SQL query per type, merged by the caller.
• DROP INDEX of a member index in SQL leaves the logical entry covering fewer types.
• Schema introspection lists the per-type indexes, not the one logical row SHOW INDEXES gives in Cypher.

Goal

Data created through the Cypher statement is searchable by the same name from SQL.

Options

1. Resolve the logical name in SEARCH_INDEX, SEARCH_FIELDS, vector.neighbors and the other index-name functions: fan out to the member indexes and merge. Needs the resolver moved out of opencypher into a place both engines use. Full-text scores merge by raw score (BM25 statistics are per per-type index), as in Cypher.
2. Do not add a SQL form, and document the supertype pattern for SQL users: an index on a common supertype is polymorphic, has one set of BM25 statistics, and needs no logical name.
3. Make a SQL DROP INDEX of a member keep the logical entry consistent (clear or refuse), independent of 1 and 2.

Decide between 1 and 2 first; 3 is wanted either way.