A postgresql column which is non-null and unique is treated with
the same optimisations as a primary key, so we have no need to
try and recreate the `id` as the primary key.
No read paths are impacted by this change, and the database will
automatically create an ID for each insert. Thus, no code needs to
change.
