together.

Switch/nrt - Properly track usage of CAU and RDMA resources with multiple
tasks per compute node. Previous logic would allocate resources once per
task and then deallocate once per node, leaking CMA and RDMA resources
and preventing their use by future jobs.

If a job is held, then only release it with the "scontrol release <jobid>"
command rather than a simple reset of the job's priority. This is needed to
support job arrays better. Otherwise a priority reset of a job array
would free all requeued/held jobs from that job array rather than
leaving them held.

If a job's priority is set non-zero then always clear the
JOB_SPECIAL_EXIT job state flag, not only when the prior
state is HELD_USER or HELD.
I'm not sure how the job could have cleared the HELD state
and changed to NO_REASON, but this would fix the problem.
bug 760

Modify slurmd to keep track of which jobs have already been launched.
It the launch is complete, then process suspend requests immediately.
Previously the suspend request was always delayed by 1 second, which
adversely impacts gang scheduling performance. If the job can't be
found (say after a slurmd restart), then delay the suspend by up
to 3 seconds, but only once.

Change the integer to hex function to support 32-bit unsigned
integers and exit on systems with more than 32 cpus per node
since Expect can not work with numbers so large.

Change the integer to hex function to support 32-bit unsigned
integers and exit on systems with more than 32 cpus per node
since Expect can not work with numbers so large.

Conflicts:
	src/slurmd/slurmstepd/fname.c

This corrects some anomalies in task_id handling. The most significant
fallout is the "%a" task ID printed for a NOT array job is now
0xfffffffe rather than 0xfffe, but this should not be used anyway.

If the job's stdout or stderr file name contain a "%A" (Job array
ID) and it is not a job array, then treat like "%j" (job ID).

If the job's stdout or stderr file name contain a "%A" (Job array
ID) and it is not a job array, then treat like "%j" (job ID).

This actually causes more problems since there are a bunch of
places where we set the reason only if it was previously NO_REASON,
but we want to treat this the same way, i.e. untested. Adding
logic everywhere to clear and reset this value seem to create
too much overhead for very little value.

Previously partition priority was only considered when used as a
component of a job's priority with the priority/multifactor plugin.
Now the partition priority is considered first, as documented,
and the job priority is considered second.
bug 764

There were a couple of cases where a job's reason
should have been set, but was not. In addition the
state_desc string should have been cleared in some
cases, but was not.

Plus clear a job's state_desc when the reason is set to NO_REASON

in 2.0 :)

See bug 757

Previously partition priority was only considered when used as a
component of a job's priority with the priority/multifactor plugin.
Now the partition priority is considered first, as documented,
and the job priority is considered second.
bug 764

Add --priority option to the salloc, sbatch and srun commands.

This code was originally put here to enforce checks to make sure jobs
didn't go over the limit.  If they didn't request the amount then we set
the limit and worked off that as if it were a request.

If we do this now we could get jobs deigned which would cancel the job at
submit with a very unrelated note as to why the job failed.

Since we now check this these limits after the node selection this isn't
needed.