Add per-cluster deployment health signals (hasIssues, isProvisioning)

[gianlucam76]

The /capiclusters and /sveltosclusters endpoints now include two new boolean fields on each cluster:

• hasIssues — true when at least one ClusterSummary has a feature in Failed or FailedNonRetriable state, or in Provisioning state with a non-empty failure message. The latter covers the Sveltos retry cycle where a retriable error moves status back to Provisioning without clearing the failure message; the message is only reset when the feature reaches Provisioned.
• isProvisioning — true when at least one ClusterSummary is actively deploying (clean Provisioning with no failure message) or waiting for a dependency (status.dependencies non-empty). Always false when hasIssues is true, so the two signals are mutually exclusive from the consumer's perspective.

Both signals are maintained as secondary in-memory indexes (clusterWithIssues and clusterProvisioning, each a map[string]*libsveltosset.Set) updated by the existing ClusterSummary watcher. No per-request scanning — each cluster lookup is O(1). Entries are created lazily and deleted when the set empties, keeping memory cost proportional to clusters with active state rather than total ClusterSummary count.