E2E demo: Full observability → investigation → fix PR pipeline

[nomadicmehul]

Summary

Create an end-to-end demo scenario that showcases the complete Phase 1-3 pipeline: alert from Grafana → agent queries logs/traces/metrics → RCA with full evidence → auto-generated fix PR.

Why This Matters

This is the "wow" demo for meetups, conferences, and README recordings. It shows TheNightOps going from alert to fix PR autonomously.

Scenario

1. Deploy demo app with a memory leak (existing memory-leak scenario)
2. Grafana alert fires → TheNightOps receives webhook
3. Agent checks K8s (OOMKill detected) → Queries Grafana for memory trends → Queries Sentry for error stack trace
4. Agent builds RCA: "Memory leak in BatchProcessor introduced in v2.4.1"
5. Agent generates PR: bump memory limit + link to code fix recommendation
6. Dashboard shows full investigation timeline with observability evidence

Acceptance Criteria

• Works with scripts/demo-gke.sh or local mode
• Demo Grafana instance with pre-configured dashboard + alert rule
• Full pipeline runs in < 5 minutes
• Dashboard captures every step visually
• Screen recording for README / conference talks