dv changes. add propagation of disconnect messages, dv updates on peer direct connections. Result of those two changes should be a much better learned topology for all peers. This was periodic gossip is really only to stop timeouts from removing peers. Also changed dv_api to wait for message queuing by dv service before calling transport transmit continuation, so that addresses that are no longer available can be caught by transport service. Perhaps some other stuff I have forgotten about. Should be kinda sorta working, need to make better testcases and do more debugging as errors inevitably pop up