Add sprio --local and --sibling options for use in federation of clusters.

Move the function to get job priorities from src/sprio to src/api
  so that we can handle the threading required for federated clusters
  in a common location in the event this logic is used from other
  places in the future (e.g. sview). No fundamental change in logic.
  The logic and definitions were just moved.

Add sprio -p/--partition option to filter jobs by partition name.
Add partition name to job priority factor response message.

Modify backfill logic so that bf_max_job_user, bf_max_job_part and
    bf_max_job_user_part options can all be used independently of each other.

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.

The problem will be solved by using per-partition QOS and setting
MaxJobPer[User|Account] on that QOS.
bug 3668

This was a regression in commit 1967d512.

Bug 3650

This fixes test 7.17 to work with commit a152196f

And mark thread local variables static to limit scope.

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.