helm-controller: detect drift in cluster

[jasaltvik]

Sometimes a resource managed by the Helm Operator (through a HelmRelease) is deleted by mistake (e.g. with kubectl). It would be nice if the Helm Operator could detect this, and reconcile the cluster state with the contents of the Helm release at the next sync interval.

Slack discussion: https://cloud-native.slack.com/archives/CLAJ40HV3/p1597310105111800

[hiddeco]

An issue related to this can be found here: fluxcd/helm-controller#186, but without any new progress to share.