Refactor runtime artifact-gate validation below complexity budgets

Context

The ratcheted quality gates introduced by #997 measure packages/runtime/src/artifact-gates.ts at 8,934 nonblank/noncomment lines (target: 500). Its largest hotspots include:

• verifyCurrentCampaignTimeoutEvidence: cyclomatic 122 and cognitive 116 (target: 15)
• verifyCoverageProductionInventory: cyclomatic 83 and cognitive 122 (target: 15)
• verifyFinalReportImplementationCoverage: cognitive 93 (target: 15)

The baseline prevents further growth, but the current structure is difficult to review and test safely.

Proposed work

• Extract cohesive validators by artifact/evidence domain behind small typed interfaces.
• Preserve fail-closed behavior and diagnostic wording with characterization tests.
• Avoid a file swarm or generic helper layer that merely moves branches without reducing complexity.
• Lower the committed metric baseline in each refactoring PR.

Acceptance criteria

• No extracted production function exceeds cyclomatic or cognitive complexity 30 as an intermediate ceiling.
• artifact-gates.ts falls below 6,000 measured lines in the first milestone.
• Existing artifact-gate tests and release validation remain green.
• .quality/metric-baseline.json is pruned for every resolved hotspot.

Follow-up to #997.

Start with packages/runtime/src/artifact-gates.ts and inspect verifyCurrentCampaignTimeoutEvidence, verifyCoverageProductionInventory, and verifyFinalReportImplementationCoverage. Read the existing artifact-gate tests, release validation, and .quality/metric-baseline.json before choosing cohesive validator boundaries. Done means preserving fail-closed behavior and diagnostic wording, keeping extracted functions below the stated ceiling, reducing the file below 6,000 measured lines, and pruning resolved baseline entries.