Uniqueness is only based on the id from now on.
Also adds the 'id' column to the Cassandra schema (it was already present in postgresql's schema)
Depends on D5029.
the code in D5029 MUST NOT be released and deployed as the same time as this one (see comment in the migration)