fix(kuma-dp) remove the datplane of any completed Pod

[nickolaev]

Summary

Since now we do support Jobs, we need to also make sure there are no offline dataplane proxies, as a result of completed jobs but still existing pods.
Delete the relevant resource and make sure it does not appear again while the containers in the pod are terminated.

Documentation

• Not needed

[jakubdyszkiewicz]

it's fix(kuma-cp)

[lobkovilya]

To be honest I don't think it's Kuma's duty to clean up Dataplanes for completed Pods. Dataplane already has an owner reference to the Pod, if Pod exists then it's fine that Dataplane exists. If Pod terminated then Dataplane should reflect it with health: ready: false and I think that what happens right now. When you create Job or CronJob there are mechanisms like successfulJobsHistoryLimit or ttlSecondsAfterFinished to clean up finished Pods

[nickolaev]

To be honest I don't think it's Kuma's duty to clean up Dataplanes for completed Pods. Dataplane already has an owner reference to the Pod, if Pod exists then it's fine that Dataplane exists. If Pod terminated then Dataplane should reflect it with health: ready: false and I think that what happens right now. When you create Job or CronJob there are mechanisms like successfulJobsHistoryLimit or ttlSecondsAfterFinished to clean up finished Pods

You are absolutely right about all that you enumerated. However, there is one small detail that is the actual reason for this PR - jobs are short-lived services and are expected to complete relatively fast. Yes, there are mechanisms to cleanup jobs, yet if they are not set, or even before they do their cleanup, one will get unhealthy data planes. For no reason, other the fact that the job is meant to just end.