Don't try to reconnect control connections indefinitely

[whatyouhide]

Right now if a control connection to a node goes down, we try to reconnect forever. This is an issue because C* cluster topologies evolve in time. For example, we have seen a case where our SaaS provider moves the nodes in our cluster around, resulting in the original seed node's control connection constantly trying to reconnect to a node that's not there (and won't be there anymore).

Control connections should "evolve" with the cluster and should follow nodes in the cluster as they go up and down. See also #212.