Scheduler crash in Airflow 2.10.5 due to TypeError in _get_ready_tis

### Apache Airflow version

Other Airflow 2 version (please specify below)

### If "Other Airflow 2 version" selected, which one?

2.10.5

### What happened?

I'm encountering an issue with Apache Airflow 2.10.5 where the scheduler crashes with a TypeError. This issue happens when it detects a zombie task and tries to process it. The scheduler logs and stack trace indicate that the error occurred due to a NoneType value for end_date in the next_retry_datetime method of TaskInstance.

Here are the relevant logs and traceback:

> [2025-06-12T01:43:19.317+0800] {scheduler_job_runner.py:2110} ERROR - Detected zombie job:
{'full_filepath': '/data/airflow/dags/analysis_hourly.py',
'processor_subdir': '/data/airflow/dags',
'msg': "{'DAG Id': 'analysis_hourly', 'Task Id': 'analysis_hourly.tableau-235976', 'Run Id': 'scheduled__2025-06-11T16:00:00+00:00', 'Hostname': 'centos-hadoop3dn-480896.intsig.internal', 'External Executor Id': '52b249a2-d690-4713-a3b3-1b2e8d72305b'}",
'simple_task_instance': SimpleTaskInstance(dag_id='analysis_hourly', task_id='analysis_hourly.tableau-235976', run_id='scheduled__2025-06-11T16:00:00+00:00', map_index=-1, start_date=datetime.datetime(2025, 6, 11, 17, 34, 33, 41895, tzinfo=Timezone('UTC')), end_date=None, try_number=1, state='running', executor=None, executor_config={}, run_as_user=None, pool='default_pool', priority_weight=2, queue='worker_03', key=TaskInstanceKey(dag_id='analysis_hourly', task_id='analysis_hourly.tableau-235976', run_id='scheduled__2025-06-11T16:00:00+00:00', try_number=1, map_index=-1)),
'task_callback_type': None}

Full traceback:

> File "/data/miniconda/envs/py311/lib/python3.11/site-packages/airflow/models/taskinstance.py", line 2601, in are_dependencies_met
for dep_status in self.get_failed_dep_statuses(dep_context=dep_context, session=session):
File "/data/miniconda/envs/py311/lib/python3.11/site-packages/airflow/models/taskinstance.py", line 2625, in get_failed_dep_statuses
for dep_status in dep.get_dep_statuses(self, session, dep_context):
File "/data/miniconda/envs/py311/lib/python3.11/site-packages/airflow/ti_deps/deps/base_ti_dep.py", line 115, in get_dep_statuses
yield from self._get_dep_statuses(ti, session, cxt)
File "/data/miniconda/envs/py311/lib/python3.11/site-packages/airflow/ti_deps/deps/not_in_retry_period_dep.py", line 48, in _get_dep_statuses
next_task_retry_date = ti.next_retry_datetime()
^^^^^^^^^^^^^^^^^^^^^^^^
File "/data/miniconda/envs/py311/lib/python3.11/site-packages/airflow/models/taskinstance.py", line 2685, in next_retry_datetime
return self.end_date + delay
~~~~~~~~~~~~~~^~~~~~~
TypeError: unsupported operand type(s) for +: 'NoneType' and 'datetime.timedelta'

From the logs, it seems that end_date in the TaskInstance object is None, causing the crash when a zombie task is being processed.

Steps to Reproduce:

1. Schedule a DAG with tasks that may encounter retries or fail conditions.
2. Observe logs where the scheduler detects zombie tasks (ERROR - Detected zombie job).
3. Scheduler crashes with the above traceback.

Expected Behavior: The scheduler should handle zombie tasks gracefully without crashing.

Actual Behavior: The scheduler crashes due to a TypeError in TaskInstance.next_retry_datetime when end_date is None.

Environment:

1. Airflow version: 2.10.5
2. Python version: 3.11
3. Database backend: MySQL 8.0
4. Executor: Celery-based worker
5. OS: CentOS7.9
6. DAG configuration: Includes retries and uses default_pool.

Additional Context:

- I suspect that the issue happens when TaskInstance.end_date is None during processing of zombie tasks.
- This only seems to occur under specific conditions, such as failed or zombie tasks.
- Link to relevant documentation mentioning zombie tasks [here](https://airflow.apache.org/docs/apache-airflow/stable/core-concepts/tasks.html#zombie-undead-tasks).
- This issue did not happen in earlier versions (e.g., 1.10.x).

Let me know if additional logs, configurations, or DAG definitions are needed to investigate further.

### What you think should happen instead?

_No response_

### How to reproduce

Unfortunately, I have not been able to clearly identify the exact steps to reproduce the issue. The problem appears sporadically in my environment under the following conditions:

A DAG is scheduled with tasks that include retries in their configuration.
A task runs and encounters some form of failure, potentially causing zombie tasks to appear.
The scheduler detects zombie tasks (ERROR - Detected zombie job) and subsequently crashes due to a TypeError.

### Operating System

NAME="CentOS Linux" VERSION="7 (Core)" ID="centos" ID_LIKE="rhel fedora" VERSION_ID="7" PRETTY_NAME="CentOS Linux 7 (Core)" ANSI_COLOR="0;31" CPE_NAME="cpe:/o:centos:centos:7" HOME_URL="https://www.centos.org/" BUG_REPORT_URL="https://bugs.centos.org/" CENTOS_MANTISBT_PROJECT="CentOS-7" CENTOS_MANTISBT_PROJECT_VERSION="7" REDHAT_SUPPORT_PRODUCT="centos" REDHAT_SUPPORT_PRODUCT_VERSION="7"

### Versions of Apache Airflow Providers

_No response_

### Deployment

Virtualenv installation

### Deployment details

_No response_

### 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 with airflow/models/taskinstance.py at next_retry_datetime and airflow/ti_deps/deps/not_in_retry_period_dep.py, then trace the zombie handling path around scheduler_job_runner.py:2110. Reproduce or isolate the None end_date state described in the traceback. Done means zombie-task processing no longer crashes the scheduler and the behavior is covered by a regression test.

Written by the indexing model from the issue text.