[18.0] queue_job: job failed to be requeued or cancelled if function is removed while job not yet done

[RyanTranAVB]

Module

queue_job

Describe the bug

If a job having its function (job.method_name) removed, while being pending/enqueued or failed (maybe, I haven't double checked but I think loading failed jobs during initialization), the job runner will stuck trying to requeue the job as failing to load the job. Trying to cancel to unlink the job from UI will also fail due to the same reason.

To Reproduce

Affected versions: 18.0 but properly other versions too.

Steps to reproduce the behavior: Explained in the description.

Expected behavior
I don't really know whether this is truly a bug, or it's missing a reliable way to prevent this from happening, maybe it's a rule of thumb to not having function removed if it's used in queue job, but since function can be enqueue anywhere using with_delay or delayable(), it's hard to tell which functions are used with queue job. Unless we always define job as job function to allow enqueueing it.

A way to prevent this issue from happening without enforcing how to implement jobs or control the development is to not loading the job if we are just setting the state of it (requeue, cancel or set done).

Additional context
Below is where it fails, since method_name is removed, getattr simply break

  File "/opt/odoo/src/queue/queue_job/models/queue_job.py", line 309, in _change_job_state
    job_ = Job.load(record.env, record.uuid)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/odoo/src/queue/queue_job/job.py", line 213, in load
    return cls._load_from_db_record(stored)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/odoo/src/queue/queue_job/job.py", line 288, in _load_from_db_record
    method = getattr(recordset, method_name)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^