Add governance benchmark substrate

[pengfei-threemoonslab]

Summary

• Promote benchmark/agent-pr-governance/cases.yaml to v0.2 with executable, catalog-only, and external-evidence case statuses.
• Add an experimental governance benchmark substrate that materializes base/head repos, runs the real verifier, builds capability locks, diffs them, and evaluates capability semantic expectations.
• Add schema-safe benchmark models, generated catalog/result schemas, docs, and focused tests.

Boundaries

• No report_schema_version bump; reports remain 0.23.
• Capability lock export remains 0.1; capability lock diff remains 0.2.
• No new public report field, verify integration, GitHub Action output, policy-pack behavior, or second release gate.
• release_decision.decision remains the only release gate.

Validation

• python -m pytest tests/test_governance_benchmark.py tests/test_capability_domain.py tests/test_capability_lock.py tests/test_capability_delta.py tests/test_verifier_scenarios.py -q
• python scripts/run_governance_benchmark.py --catalog benchmark/agent-pr-governance/cases.yaml --json
• python scripts/generate_schemas.py --check
• ruff check .
• python -m pytest tests/test_schema_boundaries.py tests/test_public_surface_contract.py -q
• python -m pytest tests/test_docs_links.py -q