Consider errors from htcondor scheduler

[fraimondo]

Under certain cirscunstances, this error might appear:

2024-10-28 01:02:39,583 - joblib_htcondor.backend - ERROR - Traceback (most recent call last):
  File "/home/fraimondo/dev/tbox/joblib-htcondor/joblib_htcondor/backend.py", line 907, in _watcher
    self._client.submit(
  File "/home/fraimondo/miniconda3/envs/julearn/lib/python3.12/site-packages/htcondor2/_schedd.py", line 454, in submit
    return _schedd_submit(self._addr, real._handle, count, spool)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
OSError: Failed to connect to schedd.

Currently, if this happens, then the job is not submitted but also removed from the backend queue, resulting in joblib never finishing as at least one future will not have a result.

This PR fixes this issue by:

1. Surrounding the _client.submit call in a try/except block.
2. Re-adding the job back to the backend queue if this call fails.
3. Deleting the pickled job file/
4. Showing an error in the log.
5. Waiting 1 second before continuing, to avoid spamming the scheduler that might be overwhelmed.