multinode demo

[aeyakovenko]

• at GPU speeds with 2 nodes I am seeing more window drops. still to many retransmit asks. I think the solution is to do a pipeline for this stage as well. pull all the crdt requests from the socket. TBD: reply to all of them, coalescing any duplicates.
• check single node perf again
• seems to take forever to for the crdt to discover the other nodes. need listener to use a streamer interface, and we should make the streamer threads high priority. same issue as above. (crdt pipeline #273)
• get the demo working on sagan/newton (checkout the readme; try those IPs in the scripts Aws5 #261 )
• update readme (Aws5 #261)
• Aws4 #254 merged
• logs #258 merged. this also changes event verify to call transaction verify plan
• window repair backoff. (the requests for a repair need control flow, they basically cause more packets to be sent, and duplicates, which cause more drops)
• check single node perf, print tps out of the servers. (fixed with bank sig verify bug, Aws5 #261)