Fixes: #17950 - Handle InvalidJobOperation error in job enqueueing test

[bctiemann]

Fixes: #17950

This alters the Job.delete() method to wrap job.cancel() in a try-except which handles the condition where the inner get_status call fails to find a non-null status for the job in Redis. This should fix a recurring intermittent test failure in test_enqueue_once_after_enqueue.

My investigation showed that whereas the job UUID key must be in Redis in order for fetch_job to work (which occurs in delete prior to the cancel call), it is occasionally the case that a call to self.connection.hget(self.key, 'status') returns a null value, leading to the exception being raised in rq/job.py#L400. This never occurs locally, and may be a side effect of concurrency within the CI environment.

The risk of this fix is low, I think, because the job.cancel is only being used in order to clear out a job from the cache during a delete operation; leaving a job in undeleted or partially-deleted state (particularly specifically in CI) is unlikely to have negative side effects.

[jeremystretch]

Thanks @bctiemann!