feat(troubleshoot): realistic incident recovery scenarios and deterministic verification

[howlcipher]

Objective

Implement realistic troubleshooting scenarios with deterministic recovery verification in DevOps-Learn-by-Doing.

Summary of Changes

• Core Troubleshooting Scenarios:
  • port_conflict: Port binding collision (EADDRINUSE). Distinguishes host port collisions from container internal errors, verifies socket reallocation and /health reachability.
  • missing_config: Missing required environment configuration (fail-fast startup). Distinguishes container launch from initialization crash, verifies environment injection and valid startup.
  • health_check_failure: Degraded internal dependency check (process alive != service healthy). Distinguishes running container process from degraded HTTP 503 probe response, verifies healthy state resolution.
  • resource_limit: Bounded container memory limit constraint triggering exit code 137 (SIGKILL/OOMKilled) without application traceback. Teaches OOM diagnosis and adequate memory threshold allocation.
• Scenario Lifecycle & Engine:
  • Enforces explicit lifecycle: SETUP -> INJECT -> OBSERVE -> EXPLAIN -> REMEDIATE -> VERIFY -> CLEANUP.
  • Distinguishes factual observations from interpretive hypotheses.
  • Progressive assistance (Levels 0-4) provides evidence-first learning without prematurely leaking remediation.
  • Deterministic recovery verification checks actual post-remediation system health.
  • Guaranteed finally teardown cleans up sockets, containers, and test state.
• CLI & Workflows:
  • devops-learn troubleshoot list, devops-learn troubleshoot run <scenario>, and devops-learn troubleshoot doctor.
  • Honest reporting of LIVE VERIFIED vs SIMULATED / TESTED.
• Quality & Falsification:
  • Comprehensive negative tests and falsification suite proving that wrong remediations, empty inputs, degraded dependencies, and undersized memory limits fail verification.
  • All 170 unit and integration tests passing. Full lint and type check suite clean.