Support selecting dataset sets on dataset processes

https://redmine.teklia.com/issues/5766

A new sets attribute should be added to the ProcessDataset M2M to allow users to specify one or more sets within a dataset that should be used in a dataset process. This should be an array, just like Dataset.sets.

A data migration should make all the existing ProcessDataset instances use all of the sets of the datasets, since we currently use all of the sets all of the time.

ListProcessDatasets must be updated to no longer just return the Dataset but the ProcessDataset: {"dataset": { ... }, "sets": [...]}. This will affect the frontend and base-worker, but Yoann told me not to care about it.

CreateProcessDataset must accept an optional sets attribute. When it isn't set, it defaults to all of the dataset's sets. It must validate that the sets belong to Dataset.sets.

Two new endpoints must be added, UpdateProcessDataset and PartialUpdateProcessDataset, to allow editing only the sets attribute. The same validation should also apply.

Nice to have: [Partial]UpdateDataset should either prevent renaming or removing a set when there are ProcessDataset rows that use that set, or automatically update the ProcessDataset to match the changes.