Added support for job email triggers: TIME_LIMIT, TIME_LIMIT_90 (reached
90% of time limit), TIME_LIMIT_80 (reached 80% of time limit). Applies to
salloc, sbatch and srun commands.

Add small delay to a test so that accounting records can get propagated
to the slurmdbd before we use sacct to view them.

salloc.1 sbatch.1 srun.1 nonstop.conf.5 slurm.conf.5

Per request from David Gloe (Cray)

Added CpuFreqDef configuration parameter in slurm.conf to specify the
default CPU frequency and governor to be set at job end.

Don't put jobs that are not runnable on submit into Bad Constraint state,
reject them.  Only do that if the configuration changes after the job
was accepted.

Commit c800b2d8
changed some logging from verbose to debug and this test needed
corresponding changes.
There were some other problems in the test that I also fixed.

Conflicts:
	src/slurmctld/job_mgr.c

failure as well as interactive user.

epilog complete message comes from that node do not process the batch step
information since the job has already been requeued because the epilog
script running isn't guaranteed in this situation.

start of requeued jobs.

It is only used for Hadoop and that code is still not fully fleshed out

Expand from 8 to 16 bits for more than 256 sizes

Conflicts:
	src/slurmctld/job_mgr.c

sched/backfill - Fix anomaly that could result in jobs being scheduled out
of order.
bug 911

Without this a batch script could be bound to a single CPU rather
than the full job allocation. This is a revision to commit
399cb897
without which batch scripts has no binding to CPUs via hwloc

squeue and scontrol to report a job's "shared" value based upon partition
options rather than reporting "unknown" if job submission does not use
--exclusive or --shared option.
bug 939