fix(ddl): scan SQL code, not literals, in the query-function router

[EnRaiha]

Why

query_functions::router::try_dispatch picked a query function with
sql.to_uppercase().contains(...) over the raw statement, so any keyword
inside a string literal, quoted identifier, or comment hijacked the statement.
Any statement that does not parse as typed DDL and whose text contains one of
six substrings failed with a misleading 42601, found while replaying a
knowledge-graph backfill:

SQL	before
INSERT INTO kg (id) VALUES ('tmp_verify_balance_x')	VERIFY_BALANCE requires (collection, column)
INSERT INTO kg (id) VALUES ('tmp_plain_x')	OK
INSERT INTO kg (id) VALUES ('tmp_z', 'verify_balance') (keyword in a label)	hijacked
INSERT INTO kg { id: 'tmp_verify_balance_y', ... } (brace form)	OK

Issue: #349.

What

• add scan_code: blanks '...' (with '' escapes), "...", --, and
  /* */ before the keyword checks; all other characters pass through verbatim
• extract recognized_function (the routing decision) so it is testable;
  try_dispatch matches on it; recognition order unchanged; a real
  SELECT VERIFY_BALANCE(...) still routes

Tests

• unit tests in query_functions::router now pin the routing decision
  (recognized_function): literal, escaped quote, quoted identifier, line and
  block comment, token fusion, first-keyword order, and real calls
• wire regression in nodedb/tests/wire/cases/router_misroute_literals.rs
  (value_carrying_verify_balance_does_not_misroute): a parenthesised INSERT
  whose value carries the token stores verbatim — the brace form is intercepted
  before the router, so it cannot guard this — and the anchored call still
  reaches the function arm
• local: cargo test -p nodedb --lib query_functions::router -> 9 passed;
  cargo test -p nodedb --test wire value_carrying_verify_balance_does_not_misroute
  -> 1 passed
• CI: static gates pass; the Test Suite workflow runs only with the run-ci
  label (.github/workflows/ci.yml gates it), so it reports skipping by design

Survey with c2g on the branch: diff-impact = the decision helper and its call
site; no exported surface change.

Notes

• Scanner scope: single-quoted literals ('' escapes), quoted identifiers
  (""), -- and /* */ comments. Not handled: nested block comments,
  E-string backslash escapes, and dollar-quoted strings — pre-existing class,
  not regressions of this change.
• Follow-up, separate refactor: dedupe with the private scanners in
  planner/procedural/executor/core/sql_literal_concat.rs
  (consume_quoted_region, parse_single_quoted_literal) — requires a
  visibility change.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[farhan-syah]

Closing. The fix keeps routing by substring and adds a second quote/comment scanner beside the existing one. The correct fix routes on the parsed function name, so literals, identifiers, and comments cannot match at all. The maintainers will fix #349 in-tree.