Late vnode replies show up as errors in the logs

[evanmcc]

When a cluster is heavily loaded, sometimes a vnode will not respond within the client's receive timeout. In webmachine this isn't a problem, since the receiving pid is dead after the request is finished and the message is dropped. But since the pbc interface calls gets from a long-running pid, the gen_server must occasionally deal with these messages when it gets them.

So:

• add a non-error handle_info clause for vnode replies that reach this process after a get's receive timeout has expired.
• log those as info with a clearer message

[evanmcc]

cc @jtuple @kellymclaughlin

[evanmcc]

This is a partial fix for issue #16, I suppose.

[seancribbs]

@evanmcc This seems to special-case get FSM replies, but is also stale. OK to close?

[evanmcc]

Yes, close it, it isn't the right fix.