Move will-run test for multiple clusters from the sbatch code into the API
so that it can be used with DRMAA.

especially when in memory debug mode

helgrind found the locks here were being make in reverse order
after the initial locks at the top of the function.

This reverts commit a7ad186e0f2b643c4e2bf302bc18e892d499cd4a.

helgrind found that the locks were taken in reverse order in some
places. This makes the ordering of locks consistent.

    an maintenance reservation that is not active yet.

This should result in better performance

problem with brackets and left unused variable

These changes do not effect operation, but eliminate some false
positive errors from helgrind, namely testing some variables values
that are sometimes protected by locks and other times not protected.

This fixes helgrind warning

None of these changes should really matter, but wrap some reads
of the agent thread count with mutex to prevent helgrind reporting
race conditions.

Conflicts:
	src/slurmctld/node_scheduler.c

Conflicts:
	src/slurmctld/node_scheduler.c
	src/slurmctld/read_config.c

work "partition".

reference bug 1246

always and it scares people.

Bug 1245

is requested.

This is a re-factor of commit e5635a76 related to bug 1148 to handle
the cases where a job could run, but an error was given when selecting
the nodes.

lock was locked outside of the function or not.  This also fixes a race
condition when adding a QOS and planning on using it right away when the
controller is busy with previous requests.

from slurm.h

tools.  This was fixed in 14.11, so this code should be all no-op there,
but just to be safe we removed it here.

PerCPU.

Before it wasn't taking into account if the user was requesting per node
memory or the job was told it needed to use less than the node allowed.