feat(chart): web/worker reliability — PDB, configurable probes, anti-affinity, HPA v2

[anchapin]

Summary

Adds Kubernetes reliability features to the web and worker deployments: PodDisruptionBudgets, configurable probes, pod anti-affinity, topology spread, HPA v2 with behavior controls.

Changes

Web Deployment

• PodDisruptionBudget (web-pdb.yaml): prevents all web pods being disrupted simultaneously
• Configurable probes: readiness/liveness/startup probe paths, delays, timeouts, thresholds all values-driven
• Pod anti-affinity: preferred hostname spread (optional, web.podAntiAffinity.enabled)
• Topology spread: zone spread with ScheduleAnyway (optional, web.topologySpread.enabled)
• Lifecycle preStop: configurable sleep before SIGTERM for graceful drain
• Replica count configurable via web.replicaCount

Web HPA

• Upgraded from autoscaling/v1 to autoscaling/v2
• Added web_hpa.enabled flag (default: true)
• Added scaleUp/scaleDown behavior blocks (stabilization windows, policies)

Worker Deployment

• PodDisruptionBudget (worker-pdb.yaml)
• Configurable QUEUES via worker.queues
• Moved cluster-autoscaler.kubernetes.io/safe-to-evict annotation to pod spec (was incorrectly on Deployment)
• terminationGracePeriodSeconds increased from 180 → 300

Values Added

web.replicaCount, web.terminationGracePeriodSeconds
web.readinessProbe.{path,initialDelaySeconds,periodSeconds,timeoutSeconds,failureThreshold}
web.livenessProbe.*, web.startupProbe.*
web.lifecycle.preStopSleepSeconds
web.podDisruptionBudget.{enabled,minAvailable}
web.podAntiAffinity.{enabled,type,weight,topologyKey}
web.topologySpread.{enabled,maxSkew,topologyKey,whenUnsatisfiable}
web_hpa.enabled, web_hpa.scaleUp/Down.*
worker.queues, worker.podDisruptionBudget.*
worker.container.terminationGracePeriodSeconds (180→300)