Use a UUID for the TranscriptionEntity primary key

https://redmine.teklia.com/issues/7696

We created TranscriptionEntity without specifying a primary key field, so Django created an AutoField (an integer). To be consistent with the other ML results and simplify re-importing exports, we need to replace that field with an UUIDField.

• Add the usual id = UUIDField(...) on TranscriptionEntity.

• Just like in process.0023, have the migration use a RunSQL whose state_operations include updating the type of the id field.

  • Use gen_random_uuid() to generate the new UUIDs to reset the column.

  • For databases that were created after Django 4.1, the id column will use GENERATED BY DEFAULT AS IDENTITY, a constraint that must be removed before you can change the type. You can use DROP IDENTITY IF EXISTS on the column to clean it up.

  • For databases that existed before Django 4.1, there will be a documents_transcriptionentity_id_seq sequence that you need to drop manually if it exists.

  • This migration is reversible, and easier to do using the old method before Django 4.1:

    • Create the documents_transcriptionentity_id_seq. No need to override any defaults on it.
    • Reset the field type to int, this time USING nextval('documents_transcriptionentity_id_seq'::regclass).
    • Set the same nextval call as the default value for the column, so that the sequence is used when inserting new rows.

• Update the export structure to have transcription_entity.id be a VARCHAR(37), and bump the export version.

• Update the maximum supported version in load_export. Previous versions should still be compatible.