Phase 12: publish truthful resilience lifecycle evidence

[CodeBuildder]

Closes #120

What changed

• publishes a seven-stage lifecycle for both portable and live judge incidents: Healthy, Fault injected, Detection, Decision, Human approval, Recovery, Verification
• moves the live Chaos Mesh approval prompt until after observed Argus evidence exists
• records actual live timestamps, detection time, recovery time, continuous HTTP availability, experiment ID, evidence source, replacement pod, and verification evidence
• gives the portable demo the same visual lifecycle while explicitly marking it replay/simulator and leaving availability unmeasured
• prevents either demo from declaring PASS unless Sentinel returns all seven stages and the correct availability semantics

Why

The earlier proof correlated Argus and Phoenix successfully, but judges had to reconstruct the lifecycle from two findings. It also collected fault authorization too early. This change makes the proof sequence directly auditable and preserves the human-in-the-loop claim at the moment a risky recovery action becomes eligible.

Truthfulness guarantees

• live values come from real timestamps, Kubernetes readiness, and HTTP probe samples
• simulator stages say simulated/replayed and never claim measured availability
• missing lifecycle evidence fails demo verification instead of being inferred
• live approval occurs only after observed detection

Validation

• complete make demo-platform proof: PASS with 18 entities, 16 relationships, 6 findings, and 3 correlated incidents
• bash scripts/tests/test-demo-platform.sh
• bash scripts/tests/test-demo-platform-live.sh
• shell syntax and whitespace validation

Stack

Based on Phase 11 PR #119; merge after that PR.