Possible bug with `verdi process repair`

[edan-bainglass]

@agoscinski I'm working on #6879 (rebased on main) and noticed the following issue:

(aiida-dev) (filter-active-jobs) aiida-core > verdi process repair -u aiida@localhost
Success: No inconsistencies detected between database and broker.
--- Logging error ---
Traceback (most recent call last):
  File "/home/edanb/miniforge3/envs/aiida-dev/lib/python3.12/logging/handlers.py", line 75, in emit
    logging.FileHandler.emit(self, record)
    ^^^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'FileHandler'
Call stack:
  File "/home/edanb/AiiDA/aiida-core/src/aiida/brokers/rabbitmq/broker.py", line 145, in __del__
    LOGGER.warning(f'RabbitmqBroker {self!r} was not closed explicitly.')
Message: 'RabbitmqBroker <aiida.brokers.rabbitmq.broker.RabbitmqBroker object at 0x754afb794770> was not closed explicitly.'
Arguments: ()
Warning: RabbitmqBroker <aiida.brokers.rabbitmq.broker.RabbitmqBroker object at 0x754afb794770> was not closed explicitly.

I removed my changes and retried - same issue, so unrelated/pre-existing.

[agoscinski]

So its technically a bug in our code that we do not close properly resources. But putting this detection in __del__ is a bit unstable because other python utilities like the logging also stop working at that point. A quickfix is to change it to warnings.warn instead of using the logger. I actually thought I fixed this in 0bc1b50. But it seems that a real fix is to use https://docs.python.org/3/library/weakref.html and trigger it on the finalize. This works like a "stable" __del__ but requires some more work because you need to move the resources you want to keep track of into a separate class. I think I try to do the finalize and if it takes more than half an hour, we do the quickfix. In addition we need to close the broker correctly on this verdi command. There is nothing broken in logic of the command. It is just that the message is confusing and suggests that something is broken.