Added default behavior so that when no pid/eid is given, it looks up
all the nodes that haven't reported in the last 2 hours and runs it
on those nodes.
Also added magic around ssh calls. Fork and set an alarm, if it doesn't come
back, declare the node unreachable and kill the child.
(This checkin is for checkpointing, I'm now going to add this to configure
so that it properly gets devel libdb.pm instead of hardcoded to the real one.)