Save the database in the archive and make one archive per dataset

We have to rework the way this worker handles the database. Instead of relying on configure_cache, the database will be created during process_split and saved in the archive.