[kube-prometheus-stack] Allow disabling individual default Grafana dashboards

[alien2003]

Closes #3763

Today the curated Grafana dashboards can only be toggled as a group through grafana.defaultDashboardsEnabled. Dropping a single dashboard means overriding its ConfigMap out of band, which the chart overwrites on the next upgrade.

This adds grafana.defaultDashboardsDisabled, a map keyed by dashboard name (the file name under templates/grafana/dashboards-1.14 without the .yaml suffix). A name set to true stops that dashboard's ConfigMap and its optional GrafanaDashboard CR from rendering while the rest stay:

grafana:
  defaultDashboardsDisabled:
    nodes-darwin: true
    k8s-resources-windows-cluster: true

The default is an empty map, so existing installs render the same dashboards as before. The approach mirrors the existing defaultRules.disabled opt-out for alerts. index is used instead of dot access because dashboard names contain dashes.

The dashboard templates are generated, so the guard clause was added to hack/sync_grafana_dashboards.py and applied to the rendered files; the committed files are identical to the generator output for this change.

How this was tested

• helm unittest --strict --file 'unittests/**/*.yaml' charts/kube-prometheus-stack (64 tests pass, including a new unittests/grafana/default_dashboards_disabled_test.yaml covering the default render, a disabled dashboard rendering 0 documents, an unrelated dashboard still rendering, and the GrafanaDashboard CR path).
• Verified each of the 34 generated dashboard guards matches what the updated generator emits.

Backward compatibility

No breaking change. New optional value, default {} preserves current output. Minor version bump 86.2.3 -> 86.3.0.