Forbid creating new worker results with worker version IDs

Closes #1420 (closed)

• This uses two different strategies to implement read-only forbidden fields and write-only forbidden fields:
  • On all endpoints where nobody is using the returned worker version ID, the field has been completely removed, and a ForbiddenField() is used instead. The field is never returned in the response, and using it in the request body causes an error.
  • On all endpoints where the worker version ID might still be used, which happens often with the frontend as it stores the result of the API request directly into the store, the worker version ID fields are replaced with a basic UUIDField with a ForbiddenValidator(): the worker version ID is still included in the response, but using it in the request body will cause an error. This also means we cannot document that those are read-only in OpenAPI.
• Sending a worker version ID of None cannot be detected due to how DRF validates null values before any custom validation can happen, and changing this requires way too much effort for the benefit it could give here. Since sending None implies you want to create something manual or something with a worker run ID anyway, it doesn't matter much.
• On bulk endpoints, where a worker run will be required, it is now properly marked as required in the OpenAPI docs and DRF will throw basic This field is required. errors.
• All worker_version and worker_version_id fields in the relevant endpoints are now also marked as deprecated.
• The ValidateClassification and RejectClassification endpoints were not actually relevant to this MR. These endpoints do not take in any request body, but the OpenAPI documentation was saying that they were because DRF and Spectacular were not expecting their weirdness. I just removed the useless request bodies for them.
• I added a whole bunch of assertNumQueries and other assertions in the tests I was updating as there were a lot of places where I could not properly verify what I was doing…
• ⚠ Merging this breaks the PDF text import (both from a DataFile and from S3) and the Transkribus import