(In progress)

Discourage users from using clusters with many procs/jobs -- this makes reduces the advantages of DAGMan, especially if they're using node retries.

Document special provisions in Job object for dealing with this...