ocrd network extension

[joschrew]

This are mainly only small additions to ocrd_network which I need for the slim containers.
Summary:
- Possibility to create a queue for a worker when the worker is externally managed. Previously only the processing-server was able to create a queue for a worker.
- Retry for queue creation: When a worker is started at the same time as the rabbit-mq, the rabbitmq needs up to 2 seconds to be fully available so that a queue can be created.
- allow starting a processing-server without any workers or processor-servers
- limit pydantic to version 1. Version 2 was released in June 2023 and conflicts with current models which were created for pydantic v1

[joschrew]

@kba: This should be included in Core when core is updated next time. This is actually only about enabling the processing server to be started without workers and about a worker being able to create its own queue.

[MehmedGIT]

This was already merged when I was on vacation, however, I thought we wanted to have an endpoint on the Processing Server for registering external (not deployed at start time) processing workers and processor servers. This way the additional --create-queue and --queue-connect-attempts flags for the processing worker are not needed.

[MehmedGIT]

It was decided to remove the flags. The first flag --create-queue is not needed since queue creation is idempotent. The processing worker will always try to create the queue when started. The second flag could be just an environment variable.