perf(kubernetes): Remove cluster relationship from pod kind (#3758)

Pods are currently set as having a cluster relationship; this
means that we create a cluster (and associated relationship to
the pod) for every pod in a kubernetes deployment.

This is not necessary, and doesn't match the definition of cluster
used by other providers.  These clusters and relationships are
also not used when we construct the server group model for
a Kubernetes application; we end up ignoring any cluster that
does not map to a server group (which pods do not).

This change will significantly improve the efficiency of both
the caching logic, and the logic to return server groups for
kubernetes (which now will not need to process each pod individually).

clouddriver-kubernetes/src/main/groovy/com/netflix/spinnaker/clouddriver/kubernetes/v2/description/manifest/KubernetesKind.java

@@ -82,7 +82,7 @@ public final class KubernetesKind {
   public static KubernetesKind PERSISTENT_VOLUME_CLAIM =
       new KubernetesKind("persistentVolumeClaim", KubernetesApiGroup.CORE, "pvc", true, false);
   public static KubernetesKind POD =
-      new KubernetesKind("pod", KubernetesApiGroup.CORE, "po", true, true);
+      new KubernetesKind("pod", KubernetesApiGroup.CORE, "po", true, false);
   public static KubernetesKind POD_PRESET =
       new KubernetesKind("podPreset", KubernetesApiGroup.SETTINGS_K8S_IO, null, true, false);
   public static KubernetesKind POD_SECURITY_POLICY =