Split partitioning migration helpers into separate functions, as the
creation of new schema objects should happen in a normal migration,
while the background migration should be called from a post-deploy
migration.