Ops: add canary rollout runbook and performance gates for web scaling changes

Summary

We need a repeatable canary rollout workflow for performance changes (HPA/pod spread/LB annotations/probe tuning) with clear success and rollback criteria tied to page-load outcomes.

Proposed work

• Add a rollout runbook for staged web deployment changes (1 pod -> canary -> broader rollout)
• Define metric gates for promotion/rollback (total page load, error rate, timeout rate)
• Provide example Helm value overlays for conservative vs aggressive rollout
• Document post-deploy validation commands and dashboards

Acceptance criteria

• Runbook supports dry-run and staged apply
• Explicit rollback triggers are documented and testable
• Promotion checklist includes measurable performance thresholds
• Guidance aligns with current chart controls (HPA v2 behavior, PDB, anti-affinity, topology spread, LB annotations).

Start by reviewing the current Helm chart controls for HPA v2 behavior, PDB, anti-affinity, topology spread, and load-balancer annotations. Document dry-run and staged apply steps, conservative and aggressive value overlays, validation commands, dashboards, measurable promotion thresholds, and explicit rollback triggers for page-load, error, and timeout metrics.