workload clusters: fix _internal.is_upgrade

Closes #1831 (closed)

This MR changes how we compute is_upgrade.

We were using the presence of the sylva-units-status ConfigMap as a criteria to determine that sylva-units had been fully deployed once.

For workload clusters, this criteria is defeated by the fact that for workload-clusters this ConfigMap is not generated in the right cluster (it's generated in the workload cluster itself, while the intent was to have it defined in the workload cluster namespace of the mgmt cluster). This specific point is being fixed in !3244 (merged), but since it won't magically fix past deployments we need to use a different criteria than the existence of this ConfigMap.

This MR changes the criteria to instead rely on the status of the sylva-units-status Kustomization to determine if it reached reconciliation at least once.