fix: assert ParentStation in TestSearchStopsByName

[Bhup-GitHUB]

Closes #377

Summary

The "returns all fields correctly" subtest of TestSearchStopsByName in
gtfsdb/fts_queries_test.go asserted 8 of 9 fields returned by
SearchStopsByNameRow, silently skipping ParentStation. Because
fts_queries.go is hand-maintained SQL (not sqlc-generated), column
scan order can drift without detection.

• Added assert.False(t, r.ParentStation.Valid) to the existing
  "returns all fields correctly" subtest, covering the NULL case for
  stop s1 (which has no parent_station set).
• Added a new subtest "reads non-null parent_station correctly" that
  sets parent_station = 'hub1' via a raw SQL UPDATE, verifies the
  scan reads it correctly, then resets it to NULL so other subtests
  are unaffected.

Test plan

• make test passes
• make lint passes

[aaronbrethorst]

Sharp catch, Bhupesh — a silently skipped field in a hand-maintained SQL scan is exactly the kind of gap that leads to subtle regressions. Good defensive testing.

What Changed

• Added assert.False(t, r.ParentStation.Valid) to the existing "returns all fields correctly" subtest, covering the NULL case and completing coverage of all 9 SearchStopsByNameRow fields.
• Added a new "reads non-null parent_station correctly" subtest that sets parent_station via raw SQL UPDATE, asserts the scan reads it, and defers cleanup to reset to NULL for test isolation.

Review

No issues found. The test is well-structured: the defer cleanup maintains isolation for subsequent subtests, the NULL and non-NULL cases are both covered, and the approach of using a raw SQL UPDATE to test the scan without modifying shared fixture setup is pragmatic.