Add SchedulerParameters option of bf_max_job_user_part to specifiy the
    maximum number of jobs per user for any single partition. This differs from
    bf_max_job_user in that a separate counter is applied to each partition
    rather than having a single counter per user applied to all partitions.

set min cpus based on tasks if we are overcommitting.  If we do then it
breaks how we overcommit.

# Conflicts:
#	src/plugins/accounting_storage/mysql/as_mysql_convert.c

to remove active sibling from federated job.

so that it can be accessible to new function slurm_kill_job_msg.

to be able to remove specific sibling jobs from a federated job's active
siblings.

The fed_mgr was trying to only use the locks that it needed so that it
could increase parallelism but this could lead to locks. When using the
internal locking, you should get all the locks once and then release
them.

The fed tests are checking the state of each sibling on each cluster.
Since squeue/scontrol now combine the jobs from all clusters into one,
the tests need to be update to look at state of the specific cluster.

They should have been removed in 63b50688

Didn't get into prev commit: 28396232

Controls whether jobs local only to the cluster are shown or all sibling
jobs are shown in a federation.

Collapse the tests into the _store_locks and _clear_locks functions,
otherwise Coverity will complain about the assignment as part of a
test, as well as the side-effects only existing in development
builds (which is intentional).

Continuation of 418e1c69

specify --with-pmix.

Bug 3597

This makes it so you can also specify %slurm_with_pmix in .rpmmacros to specify
a non-standard pmix installation.

ex. squeue -M<cluster> was talking to the local cluster instead of
<clsuter>.
Continuation of 418e1c69

Continuation of add1f9d3. Had calls to fed_mgr_is_active() in another
branch.

About "MCSParameters=ondemand,noselect:groupA|groupB|groupC",
you can separate ondemand,noselect and groupA|groupB|groupC into 2
parameters like you want. The first part is common for mcs plugin and
the second part is only for mcs/group, so it could be logical to put them
into 2 different parameters.

Bug 3428

END_TIMER was being called without a START_TIMER.
Coverity CID:165991

the slurmctld.

This reverts commit a662a226.

the slurmctld.

ACTIVE: accepts submissions and schedules
INACTIVE: doesn't accept submissions or schedules
DRAIN: schedules but doesn't accept submissions
DRAIN+REMOVE: like DRAIN but cluster removes itself from federation once
all federation jobs are gone.