Implement threaded fetch queue to simplify concurrent result processing

[wesm]

Operation::Fetch is a blocking operation and involves both IO and deserialization. There may be non-trivial computation involved with a retrieved row set, so we should enable the fetching and processing to be carried out concurrently in a simple producer-consumer fashion. I don't think this needs to be complicated -- a "spooler" class with a fixed-depth queue (generally length = 1) and a mutex should suffice, with the ability to communicate EOS or error