Add programmatic support for monitoring health of DCP client

Answer questions like:

• Which nodes is the client currently connected to? Which nodes is it attempting to (and failing) to connect to? How long has the failed connection attempt been going on for?

• Which partitions is it currently streaming from? Which partitions is it trying to re-establish a stream from? How long has the failed stream attempt been going on for?

Consider adding a new callback interface that is notified whenever a connection health event occurs.