GKEStartJobOperator / KubernetesJobOperator fail to delete pod after it ends

### Apache Airflow Provider(s)

cncf-kubernetes

### Versions of Apache Airflow Providers

First seen in
apache-airflow-providers-cncf-kubernetes==10.1.0
Still seeing in
apache-airflow-providers-cncf-kubernetes==10.5.0

I didn't have the issue in 8.3.2

I can't test more recent versions due to versions constraints

### Apache Airflow version

2.10.5

### Operating System

cloud composer

### Deployment

Google Cloud Composer

### Deployment details

_No response_

### What happened

All my task end with
```
[2025-09-02, 15:07:22 EDT] {pod_manager.py:698} INFO - Pod PODNAME has phase Running
[2025-09-02, 15:07:24 EDT] {pod.py:1079} INFO - Skipping deleting pod: PODNAME
```

when the pod correctly succeeds in kubernetes around the same time

```
State: Terminated
Reason: Completed
Exit Code: 0
Started: Tue, 02 Sep 2025 15:05:37 -0400
Finished: Tue, 02 Sep 2025 15:07:21 -0400
```
I always use `OnFinishAction.DELETE_SUCCEEDED_POD`

### What you think should happen instead

Should delete the pod per `OnFinishAction.DELETE_SUCCEEDED_POD`

### How to reproduce

I have this on any task. I use `GKEStartPodOperator` and run python scripts through it. For example:
```
GKEStartPodOperator(
image="bash",
cmds=["/usr/local/bin/bash", "-c"],
arguments=[
"echo",
"hi",
],
)
```

### Anything else

_No response_

### Are you willing to submit PR?

- [ ] Yes I am willing to submit a PR!

### Code of Conduct

- [x] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)

Start by reproducing the pod-cleanup behavior with GKEStartPodOperator or KubernetesJobOperator and the DELETE_SUCCEEDED_POD setting, using the reported provider versions. Trace the operator and pod-manager logs around the transition from Running to Completed. Done means a successfully completed pod is deleted as requested, with coverage for the reported regression.

Written by the indexing model from the issue text.