I'm curious about the relational-ness of FaunaDB. e.g. How do you efficiently maintain integrity of foreign key constraints across the entire system? How fast and consistent are secondary indexes?