Implement gettransaction RPC event queue

[wtogami]

A rescan with a large quantity of wallet transactions or a massive flood of new transactions could trigger many near-simultaneous walletnotify events, which in turn would get stuck for a while with many parallel RPC calls waiting for bitcoind.

This is nowhere near fatal to Baron 0.9.3. I suspect the logs would be a mess, but Baron handles this just fine and will eventually recover on its own even if random gettransaction RPC calls fail due to the later listsinceblock that discovers previously missed 0-conf transactions.

It would be technically best if walletnotify added the txid's to a queue without delay, and a separate event loop processed each txid one at a time.