test_falsification_shining_codex.py asserts a config filename views-models retired #489

What

TestF1ConfigFiles in tests/test_falsification_shining_codex.py asserts that
views-models/models/shining_codex/configs/config_deployment.py exists. It does not. Two tests
fail locally:

FAILED TestF1ConfigFiles::test_config_deployment_exists
FAILED TestF1ConfigFiles::test_all_required_configs_exist

Why

views-models commit 3f2d98ec — "feat(#490): config_maturity.py for the 31 cm r2darts2 models —
only the 38 stepshifter sources remain on the legacy vocabulary (#496)" — replaced
config_deployment.py with config_maturity.py. Deliberate vocabulary migration on their side.
Our test still encodes the old name.

Why nobody noticed

The class is @pytest.mark.skipif(not _SHINING_CODEX.exists()), and _SHINING_CODEX is a
relative sibling path (parents[1].parent / "views-models/models/shining_codex"). In CI the
sibling repo is absent, so all four tests skip and the job is green. The assertion only runs on a
machine that has both repos checked out side by side.

So this is a test that can only fail on a developer's laptop, and only for reasons that live in
another repository. That is the more interesting finding than the stale filename.

Suggested fix

Two separate decisions:

1. The filename. Either accept config_maturity.py as the replacement, or accept either name.
   Worth checking with views-models which is intended for shining_codex specifically — chore: sync main back into development after v1.14.0 #496 says
   only the 31 r2darts2 models moved and 38 stepshifter sources stayed on the legacy vocabulary, so
   whether shining_codex is in the moved set needs confirming rather than assuming.
2. Whether this test should exist here at all. A cross-repo filesystem assertion against a
   relative sibling path is green in CI, red on a laptop, and silent about which. If the contract
   matters it belongs in views-models' own suite; if it does not, this belongs in a post-mortem
   note rather than a test.

Found while running the full suite for #487 (ACLED cursor migration). Unrelated to that change —
confirmed by the other repo's git history — so it was not fixed there.