[AK-383] E2E test: Lando's full multi-tenant deployment flow

[robinbraemer]

Migrated from Linear: AK-383

GitHub is the canonical source of truth. The Linear issue is marked as a duplicate.

---

Why

Lando's actual stack is: Python backend + Node.js mini-server + React Native frontend + PostgreSQL — each deployed per hospital client. Before onboarding Matthias, we must prove this works end-to-end through CNAP, including multi-tenant isolation.

Description

Simulate Lando's real deployment pattern on a test cluster:

1. Create a multi-chart product representing Lando's stack
2. Deploy 2+ installs (simulating 2 hospital clients)
3. Verify each install is isolated (separate namespace, DB, domain)
4. Verify private registry pull works (Lando uses private GHCR images)

Acceptance Criteria

• Multi-chart product created (backend + frontend + DB)
• Install chore(deps): update all non-major dependencies #1 deployed — all pods healthy
• Install chore(deps): update actions/checkout action to v6 #2 deployed — all pods healthy, separate namespace
• Private GHCR images pulled successfully via registry proxy
• Each install has its own database (no data leaking)
• Each install reachable at its own subdomain
• Uninstall cleans up all resources

Out of Scope

• Actual Lando Docker images (use test images that mimic the pattern)
• Production cluster — use dev/test cluster