handle non zero deletionTimestamp for child objects.

[f41gh7]

Currently, if by mistake ServiceAccount ( or any other object) was deleted, operator doesn't remove finalizer for it. Since parent object wasn't marked as deleted. It blocks kubernetes control-plain operations.

Proposed solution:

1. check for deletion timestamp.
2. if timestamp is not zero, remove finalizer and raise an error
3. object must be recreated and the next reconcile loop.

[jmleddy]

Here's the reproducer if needed:

james@M-D19FP4G4QL ~/src/kube-resources $ kubectl get sa vmagent-az1
NAME          SECRETS   AGE
vmagent-az1   0         95d
james@M-D19FP4G4QL ~/src/kube-resources $ kubectl get sa vmagent-az1 -o yaml
apiVersion: v1
kind: ServiceAccount
metadata:
  creationTimestamp: "2024-02-09T15:55:25Z"
  finalizers:
  - apps.victoriametrics.com/finalizer
<snip>
james@M-D19FP4G4QL ~/src/kube-resources $ kubectl delete sa vmagent-az1
serviceaccount "vmagent-az1" deleted
^Z
[1]+  Stopped                 kubectl delete sa vmagent-az1
james@M-D19FP4G4QL ~/src/kube-resources $ kubectl get sa vmagent-az1  -o yaml
apiVersion: v1
kind: ServiceAccount
metadata:
  creationTimestamp: "2024-02-09T15:55:25Z"
  deletionGracePeriodSeconds: 0
  deletionTimestamp: "2024-05-15T12:56:27Z"
  finalizers:
  - apps.victoriametrics.com/finalizer
<snip
james@M-D19FP4G4QL ~/src/kube-resources $ kubectl get pods
NAME            READY   STATUS    RESTARTS   AGE
vmagent-az1-0   3/3     Running   0          6d15h
vmagent-az2-0   3/3     Running   0          6d15h
james@M-D19FP4G4QL ~/src/kube-resources $ kubectl rollout restart sts vmagent-az1
statefulset.apps/vmagent-az1 restarted
james@M-D19FP4G4QL ~/src/kube-resources $ kubectl get pods
NAME            READY   STATUS               RESTARTS      AGE
vmagent-az1-0   1/3     PostStartHookError   0 (13s ago)   89s
vmagent-az2-0   3/3     Running              0             6d16h
james@M-D19FP4G4QL ~/src/kube-resources $

[f41gh7]

Changes was included to v0.45.0 release.