fix(metrics): metric calibration — fidelity improvements across all dimensions

[srbouffard]

Summary

Replaces semi-fake POC metric logic with deterministic, well-tested, production-grade detectors across all five dimensions. All scorers are now measurability-aware: unmeasurable signals return None and propagate to UNRATED at the engine level rather than forcing false zeros/bronze.

Changes by dimension

test_verification

• latest_build_passing now falls back to GitHub default-branch check-runs when no Allure URL is set — fixes the saml-integrator false negative
• coverage_pct / stability_pct return None (unmeasurable) instead of 0 when Allure is absent

documentation

• Adds accepted workflow family name aliases (docs-checks / vale, docs-checks / linkcheck, docs-checks / docs build)
• Adds equivalent README heading variants (Quickstart, Getting help)

substrate_compat

• Full YAML-aware workflow parser: handles matrix.include list-items, inline comments, literal/folded/quoted block scalars, heredoc skip, quoted run: scalars
• Exact juju channel matching (24/stable no longer matches juju4)
• Canonical-k8s detection scoped to actual execution contexts
• Fixes crash on defaults.run: mapping blocks

security_ssdlc

• canonical_repo_automation_registered now checks git tree of canonical/canonical-repo-automation for a per-repo config file instead of code search — scoped to canonical/* owner only

support_engagement

• avg_triage_days, avg_pr_review_days, response_coverage_rate return None when total_items < MIN_SAMPLE (5) — separates insufficient data from measured zero

Engine changes

• compute_leaf_applicability and compute_product now check required_metrics_for_scoring from dimensions.yaml; any None value returns INSUFFICIENT_DATA / UNRATED
• UNRATED dimensions excluded from overall medal min (only SCORED dimensions count)

Config/docs

• config/dimensions.yaml: adds required_metrics_for_scoring lists for all dimensions
• config/schemas/dimensions.schema.json: schema updated for new field
• docs/local-scoring.md: documents measurability vs unrated semantics
• docs/superpowers/artifacts/2026-07-23-metric-calibration-audit-postfix.json: before/after snapshot

Test coverage

• 239 tests passing (up from 209)
• Regression tests for every new detection case
• Integration test guards required_metrics_for_scoring entries are valid declared outputs