Add integration tests for replication and fault tolerance

[psFried]

This is kind of a large issue to tackle. The current integration test framework can handle interactions with a cluster, but what we really want is to test the fault tolerance of the cluster. What is needed is some framework to allow simulating various network conditions in addition to nodes shutting down and coming back up. In particular, it should be able to simulate network delays, segmentations between specific nodes, packet loss. Jepsen might be a good place to start looking here.