Make it possible to easily clean "old" entries from the `object_references` table in cassandra

For the record, the current process used for the postgres backend is as follow:

• the object_references table is partitionned (by insertion date)
• there is a cli tool (take a date as argument, typically the date of the last swh-graph export) that will delete all the tables which partition date is older than the given date
• there is a cli tool used by a cron to provision the new tables as needed (one month ahead)

This overall process should be kept as is for the cassandra backend.