fix: weight static + runtime in execution correctness scoring

[lezama]

Problem

After #22, execution correctness averages the runtime's static.score and runtime.score (each already weighted), which was the right move. But two gaps remained:

1. A hard crash could still earn partial credit. If code throws a fatal/execution error before any assertion runs, the runtime sub-score is 0 but a high static.score (pattern matches) averages back up to ~0.5. Static matching should not rescue code that never executed.
2. Applicability was read from the runtime output, not the test. A crash zeroes the runtime weight, so an output-based check would conclude "runtime doesn't apply" and score the test on static alone — masking the crash.

Fix

_score_correctness(raw, test) reads which dimensions apply from the test definition (static_checks / runtime_checks), not the runtime output, then averages the applicable, runtime-weighted sub-scores. A hard crash (fatal/execution error before any assertion runs, detected via zero runtime weight or a synthetic execution_error/fatal_error entry) forces correctness to 0.0. Code that runs but fails some assertions still earns partial credit.

Before / after (real e-hooks-001: 4 weighted static patterns + 2 runtime assertions)

Case	before	after
Registers both hooks but wrong meta key (runtime 2/2, static 2.0/3.0)	0.8334	0.8334
Works but trips a forbidden pattern (e.g. direct SQL)	0.5000	0.5000
Static perfect but code crashes before assertions	0.5000	0.0000

The first two rows match #22's behavior; the crash row is the net-new fix.

Tests

New python/tests/test_scoring.py (10 cases): static+runtime averaging, single-dimension scoring, forbidden hard-fail, crash detection (both signals), and the control case (runs but half assertions fail → keeps partial credit). Reference-solution-mode fixtures updated to define real checks so definition-based applicability is exercised. Full suite passes (33), ruff clean.

[github-actions]

The following accounts have interacted with this PR and/or linked issues. I will continue to update these lists as activity occurs. You can also manually ask me to refresh this list by adding the props-bot label.

If you're merging code through a pull request on GitHub, copy and paste the following into the bottom of the merge commit message.

Co-authored-by: lezama <migueluy@git.wordpress.org>
Co-authored-by: JasonTheAdams <jason_the_adams@git.wordpress.org>

To understand the WordPress project's expectations around crediting contributors, please review the Contributor Attribution page in the Core Handbook.

[JasonTheAdams]

The quality / LLM-judge dimension (30% of overall) is still unimplemented in the runtime — tracked separately.

Can you elaborate on this, @lezama? I thought the LLM-judge piece was removed in my previous, giant PR.

[lezama]

you're right — that note was stale, my bad. this PR was branched off trunk before #22 merged, so its base still had the old quality/judge scaffolding (hence the "30% of overall" line). it predated your removal.

just rebased onto current trunk and reframed the PR. #22 already does the static + runtime weighted averaging, so i've scoped this down to the actual net-new:

1. a hard crash forces correctness to 0.0 — if code throws before any assertion runs, a high static score shouldn't average it back up to ~0.5. static matches shouldn't rescue code that never executed.
2. applicability is read from the test definition, not the runtime output — otherwise a crash zeroes the runtime weight and the test gets scored on static alone, hiding the crash.

updated the description + the reference-solution-mode fixtures, 33 tests green. keeping it as this one focused commit 🙂