Add AI-native policy and workflow integrity contracts

[moonweave]

Summary

Phase 1 established the philosophy SSOT and additive figops-human-review/1 receipt domain. Phase 2 now makes the important boundaries executable without turning the LLM into a weaker renderer:

• policy_resolution is a canonical, deterministic policy resolver with precedence, provenance, conflict fail-closed behavior, immutable kernel invariants, explicit opt-out evidence, and a stable SHA-256 policy-set identity.
• workflow.intent (exploration, execution, review, promotion) is explicit and surface-aware. Active project render conflicts fail closed; direct CSV remains exploration; review/readiness stays read-only; exploration/review cannot become promotion-eligible.
• Project render manifests and figops_evidence/2 carry the policy context and workflow intent. The policy digest is verified inside the evidence envelope, not only in an adjacent manifest.
• v2 project responses expose compact, closed policy/workflow context while keeping the AI-native discovery budget intact. Detailed policy contents remain validated in the durable evidence contract.
• Root and packaged project templates declare workflow.intent: execution; config compatibility helpers preserve existing public imports.
• Generated MCP references, v2 measurement witness, public-release status, and architecture inventory are synchronized.

The runtime/result boundary remains unchanged: external runtime is disposable, project results are durable, no-replace promotion remains the only publication primitive, and raw/, hub_scripts/, and results/ stay distinct.

Validation

• Focused locked aggregate: 854 passed, 14 skipped, 1 warning, 11 subtests
• Full locked regression: 2383 passed, 78 skipped, 29 warnings, 104 subtests
• Locked Ruff: passed
• Locked compileall: passed
• No repo-local .venv; no baseline-update environment variables
• v2 tool surface: 14,433 bytes (<24 KB); v2 generated reference: 23,642 bytes (<64 KB)

No merge, tag, or release action is included. This remains a Draft PR for review.