Add policy simulation and change-impact analysis lab

[cdcavell]

Summary

Adds the Milestone 8 hands-on lab for policy simulation and change-impact analysis.

This change introduces an intermediate lab that teaches learners how to compare a baseline policy and a candidate policy against the same representative contexts before rollout.

The lab covers:

• Baseline versus candidate policy evaluation.
• Replaying identical deterministic contexts across both policy versions.
• Structured comparison of outcomes, reason codes, and contributing policy identities.
• Detection of newly denied operations.
• Detection of newly allowed operations.
• Detection of new acknowledgment requirements.
• Detection of new escalation requirements.
• Identification of unchanged cases and reason-only changes.
• Expected-change plans versus unexpected regressions.
• Boundary contexts and precedence effects.
• Tenant-specific overlay simulation.
• Historical versus synthetic simulation data.
• Sensitive-data considerations for replay corpora.
• Candidate policy identity and provenance.
• Rollout and rollback reasoning.
• Optional probabilistic-signal replay using captured observations.
• Machine-readable difference reports and focused regression tests.
• The limitations of simulation as evidence rather than proof of production safety.

The lab also establishes the required architectural invariant:

Simulation may evaluate policy and produce decision evidence, but it must never invoke protected execution.

The change also:

• Adds the lab to intermediate lab navigation.
• Updates the Labs index and lab counts.
• Cross-links the lab from Practical Policy Testing and Decision-Table Strategies.
• Cross-links the lab from Policy Versioning and Decision Provenance.
• Marks Policy simulation complete under issue [Learning]: Add policy simulation and change-impact lab #116.
• Marks Milestone 8 — Governance and Policy Architecture as complete.

The core distinction remains:

Testing asks whether a policy behaves as specified. Simulation additionally asks what would change if a candidate policy became authoritative.

The simulator exposes those changes but does not make the rollout decision or execute real-world side effects.

Closes#116