Applying a template on a process that already has the template's versions can cause IntegrityError

1. Create a process
2. Add a worker version on that process
3. Create a template from that process
4. Apply that template back onto the original process

This causes an IntegrityError because the endpoint tries to create a WorkerRun that violates the dataimport_id, worker_version_id unique constraint.

To discuss before implementing anything:

• Should the endpoint just delete all the existing worker runs and overwrite everything with the template?
• Should applying a template fail when there would be duplicates?
• Should applying a template ignore worker versions that already exist?