Fixed issue where if the slurmdbd was down past the MinJobAge and the start of the job hadn't been sent yet to throw it on the queue to be sent when the slurmdbd comes back up.

preemptive strike before 2.1 merge which will do this exact thing, but the code has changed locations so many conflicts will happen

    that of the slurmctld daemon: the same directory used for log files,
    SlurmdLogFile (if specified with an absolute pathname) otherwise the
    directory used to save state, SlurmdSpoolDir.

    affinity is enabled, keep track of the individual CPUs allocated to jobs
    rather than just the count of CPUs allocated (which could overcommit
    specific CPUs for running jobs).

meaningful if StateSaveLocation=/dev/null

Fix for handling correctly jobs that resize, and also reporting correct stats on a job after it finishes.

cleaner and tighter code

being freed by sched/backfill.

to provide detailed logging only if so configured and avoid a bunch of 
logging calls otherwise

for bit_rep_count (implcitly 1 for each node), also make the logic more clear

   i.e., Fixed a fair-share calculation bug in the priority/multifactor plugin

to avoid having srun generate an error that the node isn't ready for use

    preempted jobs at the time slurmctld restarts that can result in over-
    subscribing resources.

to eliminate the need for a cpu_rep_count argument and only require a single array
of CPU counts per node. Modify src/slurmctld/gang.c to eliminate the extra array
for simplicity, especially if we want to change CPU count per node on the fly.

causes a job suspend/resume with substantial overhead.

both file contents are the same, but there was inconsistent naming in the configure.ac
and Makefile.am).