Verts can take infinitely long to run

[Richard Kallos]

There should be some mechanism by which a wrek can consider a vertex as having failed if it takes too long to run. This may involve killing a process started by erlexec.

Suggested solution: When wrek starts a new wrek_vert Pid, conditionally set a timer to send timeout to Pid, and implement wrek_vert:handle_info(timeout, State) to do the right thing.