Fix CDS operator getting stuck whith too many broken pods

In case of an error in the CDS spec that prevents pods from getting to the Running state (e.g. a wrong image tag) the cds operator counts all these pods as unavailable. If a fix is pushed to the CDS spec the operator will not roll it out as the already non-running pods are above the max_missing_pods threshold. This will cause a deadlock where the user either has to increase the max_missing_pods or delete some pods manually.

To fix this we generaly delete stale pods that are not/not yet running without taking the max_missing_pods setting into account.

Fixes: #14 (closed)