feat(chart): support deploymentStrategy, lifecycle, minReadySeconds, terminationGracePeriodSeconds

[eenchevlp]

Bumps chart version to 0.18.5.

Fixes #1053

Summary

With replicas: 1 (the common configuration for this exporter), a plain
rolling restart briefly drops the only pod backing the Service, causing
Prometheus scrape gaps on every redeploy. spec.strategy,
spec.minReadySeconds, container lifecycle hooks, and
terminationGracePeriodSeconds were not configurable, so there was no way
to avoid this from the chart.

This adds four new optional, passthrough values:

• deploymentStrategy → rendered verbatim under spec.strategy (e.g.
  RollingUpdate with maxSurge: 1, maxUnavailable: 0 to guarantee a
  new pod is ready before the old one terminates)
• minReadySeconds → spec.minReadySeconds (hold the new pod as "not yet
  available" long enough for Prometheus to discover and scrape it)
• lifecycle → container lifecycle hooks (e.g. preStop: sleep 45 to
  let in-flight scrapes finish and service endpoints propagate before
  SIGTERM)
• terminationGracePeriodSeconds → pod spec, sized to cover the preStop
  delay

Together these enable gapless rolling updates for single-replica
deployments: maxUnavailable: 0 + a preStop sleep keeps the old pod
serving /metrics until the new pod is confirmed ready and endpoints have
propagated.

All four values default to unset/empty and are guarded with
{{- with }}, so existing releases render identically until explicitly
configured.

Test plan

• helm lint helm/
• ct lint --charts helm
• helm template test helm/ diffed against master with the new
  values unset — identical output (only the chart-version
  label/checksum differ, from the version bump)
• helm template test helm/ -f <values with all four set> — verified
  via yq that spec.strategy, spec.minReadySeconds,
  spec.template.spec.terminationGracePeriodSeconds, and
  containers[0].lifecycle.preStop render in the correct locations

[burningalchemist]

Hey @eenchevlp, looks good to me! Thanks for your contribution, let's ship it. 🚀