fix(chart): dashboard filters on job, not the unset service=s3proxy

[ynsta]

Summary

Every dashboard panel filtered on service="s3proxy", but the
ServiceMonitor-applied service label is the release-prefixed Service
name (e.g. service="cockpit-s3proxy"), so the filter never matched and
panels rendered "No data". job carries the same release-prefixed
value but is the label Prometheus actually exposes consistently.

This patch:

• Adds a job template variable (label_values(http_requests_total, job))
  between the existing datasource and instance variables, so the
  dashboard picks up whatever job string Prometheus emits.
• Replaces service="s3proxy" with job=~"$job" in every panel target
  and in the instance variable's query.
• Adds the same job=~"$job" filter to the s3proxy_* panels (panels 5
  and 6) so they also scope correctly in multi-release clusters.

Chart bump 1.9.1 → 1.9.2. No values changes.

See CHANGELOG.md → Chart / 1.9.2.

Test plan

• jq . charts/s3proxy/dashboards/s3proxy.json >/dev/null — valid JSON
• helm lint charts/s3proxy — passes
• helm template t charts/s3proxy --set grafanaDashboard.enabled=true --show-only templates/grafana-dashboard.yaml | grep -cF 'job=~\"$job\"' → 11 (≥ 11)
• Same render: grep -cF 'service=\"s3proxy\"' → 0 (no stale filter)
• Apply against the cluster that triggered the bug; confirm panels
  populate and the Job picker lists e.g. cockpit-s3proxy.
• CI green on chart/dashboard-job-variable.

Release flow

Merge → tag chart-v1.9.2 on main → CI publishes
ghcr.io/intrinsec/s3proxy/charts/s3proxy:1.9.2.

🤖 Generated with Claude Code