[Feature] Separate controller namespace and CRD namespaces for KubeRay-Operator Dashboard

[400Ping]

Why are these changes needed?

• In multi-tenancy deployments, the KubeRay operator usually runs in a dedicated namespace, while CRD instances (RayCluster, RayService, RayJob) live in tenant namespaces.
• The current KubeRay Operator Grafana dashboard filters controller-runtime metrics and CRD metrics with the same namespace variable, so it can’t correctly display operator metrics when operator and CRDs are in different namespaces.
• This PR introduces a new dashboard variable ControllerNamespace and switches all controller-runtime queries to use it, leaving CRD queries on the existing namespace variable.

Result: the dashboard works for both single-tenant and multi-tenant setups without breaking existing users.

Related issue number

Closes #4013

Checks

• I've made sure the tests are passing.
• Testing Strategy
  • Unit tests
  • Manual tests
  • This PR is not tested :(

[400Ping]

This CI error doesn't seem to be related to this pr?

[owenowenisme]

LGTM.
But could you please add a screenshot that demonstrates both the controller metrics and the CRD metrics being displayed correctly on the dashboard?
Your current screenshot only shows controller.
Thanks!

[400Ping]

LGTM. But could you please add a screenshot that demonstrates both the controller metrics and the CRD metrics being displayed correctly on the dashboard? Your current screenshot only shows controller. Thanks!

ok, will do it later tonight