ci: fail the build when the test suite fails

[soedirgo]

What kind of change does this PR introduce?

CI fix.

What is the current behavior?

The tests check is green even when the suite fails.

The test container runs:

bash -c "./bin/installcheck; cp ... /home/splinter/results_out/ 2>/dev/null || true"

installcheck's exit code is discarded by the ;, and the trailing || true guarantees the container exits 0. Every run reports success.

In #179, the new lint was never loaded by bin/installcheck, so its regress test errored:

test 0030_invalid_index ... ERROR:  relation "lint.0030_invalid_index" does not exist
FAILED
 1 of 29 tests failed.

What is the new behavior?

1. The container exits with installcheck's status. Artifacts are still copied out on failure, so results/regression.diffs remains available for debugging.

2. New bin/check_lints.py, wired in as a pre-commit hook (so it gates every PR via the existing pre-commit hooks workflow). Registering a lint touches five files outside lints/, and forgetting most of them fails quietly. For each lints/NNNN_<name>.sql it asserts:

1. the view is named lint."NNNN_<name>" (matches the file stem)
2. bin/installcheck loads it
3. test/sql/queries_are_unionable.sql unions it
4. docs/NNNN_*.md exists
5. test/sql/NNNN_*.sql and test/expected/NNNN_*.out exist
6. mkdocs.yaml links the doc page

Plus: no two lints share a number, and every numbered doc page is in the nav.

Run against #179's first commit, it reports exactly what was missing:

3 problem(s) found:

  - docs/0030_invalid_index.md: not listed in mkdocs.yaml; add it under `nav:` -> `Lints:`
  - lints/0030_invalid_index.sql: not covered by test/sql/queries_are_unionable.sql; add `union all select * from lint."0030_invalid_index"`
  - lints/0030_invalid_index.sql: not loaded by bin/installcheck; add `-f lints/0030*.sql` before `-d contrib_regression`