Fixes:
- Sort aliased clusters by id, then name, when inserting, rather than random processing order
- If a data type reference has been resolved by one cluster alias, don't short circuit before adding the data type to the list of cluster references for all the other aliases of that cluster