We tried to do something overly complicated with the callback rpc
timeouts here.  And they're wrong--the result is that by the time a
single callback times out, it's already too late to tell the client
(using the cb_path_down return to RENEW) that the callback is down.

Use a much shorter, simpler timeout.
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>