[dylanbstorey]

Summary

• SET n = {id: n.id, q: 1} left n with {q: 1}: bulk SET cleared n, then dropped the map value it could not store (it stores only literal map values). A SET n = map with computed values now runs in the row pipeline, which computes the map for the row first, as += already did (GQLITE-T-0479).
• A SET value with a label test (CASE WHEN n:A THEN … END, n:A) was refused ("SET value must be a literal …"), with or without RETURN: the row pipeline did not know label tests were computable. It now computes them.
• known.json: the GQLITE-T-0479 entry is removed.

Test plan

• tests/functional/109_set_map_case.sql
• Before the rebase onto fix: writes after two MATCH clauses run once per row (GQLITE-T-0477) #202: unit 1030/1030; functional; TCK 3880/3880 on SQLite 3.53 and 3.54.1; Python 1669 passed; metamorphic cover reps 2 and random seeds 1, 50001, 90001: 0 new signatures.
• After the rebase: functional, TCK 3880/3880, metamorphic cover: 0 new signatures.

Semantic coverage matrix

• tests/functional/109 covers SET = map reading the node, and label tests in SET values.

Checklist

• Tests added
• No version bump