nodewatcher WS: more robust retry system (e.g. with exponential delay)

[pmespresso]

When remote node drops connection, api retries connection until it reconnects, at least in the browser, but from a pod, "something" happens and the connection is never recreated. Current workaround (read: hack) is to drop the socket entirely and create a new socket. Better solution would to persist the current blocknumber somewhere before and retry the connection until it works.