Resources for hybrid preferences research where we learn how to route preference instances for human vs. AI feedback