Add healthchecks to monitoring unit.

Summary

We've observed on some deployments that monitoring unit was seen as Ready whereas the prometheus pod was in cashLoop.

Since it is relying on standard Helm HealtchChecks, the HelmRelease will become ready as soon as the deployment that it produces (like prometheus operator) will become Ready, but it will consider the prometheus.monitoring.coreos.com resource as Ready as soon as it is applied on the cluster since it's a custom resource.

Consequently, Prometheus errors are silently ignored.

We should add some HealtchChecks to the controlling kustomisation to check that Prometheus resource is actually Ready.