to a job, but which are not responding when the slurmctld restarts. This
 should help prevent race conditions when slurmctld crashes and we try to 
 reconver consistent state (e.g. did ResumeProgram complete or not).

fixed issue where the dbd gets a start for a job it already knew about but pushed the last roll back

  since we may be waiting for node suspend/resume processes to complete

General clean up and expansion of power save documention in slurm.conf man page plus
the powwer save web page.

    after a node has been suspended before resume it (e.g. powering it back 
    up).

also only re-wake nodes that should be awake, but are not reposnding

  which results in arithmetic operation with an invalid octal value

  error() to info().

   nodes are still booting.

   powered down nodes.

  as appropriate

  suspend_program and resume_program are defined (used to only 
  check for resume_program)

   from debug2() to info()