No change in logic. Just make error messages more uniform and complete

This takes the user to the home page rather than the download page.

Convert the data structures used to implement RequeueExit and
RequeueHoldExit from an array to a bitmap, which has far better
scalability.

This just releases the requeue_exit and requeue_exit_hold arrays
when the slurmctld shuts down so we can more easily identify any
real leaks.

This reverts commit 5bd97a2c.

Conflicts:
	slurm.spec

RequeueExit and RequeueExitHold configuration parameters modified to accept
numeric ranges. For example "RequeueExit=1,2,3,4" and "RequeueExit=1-4" are
equivalent.
bug 1549

Fix systemd paths in spec file

Systemd service files were being created in the right place /usr/lib/systemd, but the test for their presence was looking in /lib, not /usr/lib.

Conflicts:
	NEWS
	testsuite/expect/test20.12

If SchedulerParameters value of bf_min_age_reserve is configured, then
a newly submitted job can start immediately even if there is a higher
priority non-runnable job which has been waiting for less time than
bf_min_age_reserve.

If SchedulerParameters value of bf_min_age_reserve is configured, then
a newly submitted job can start immediately even if there is a higher
priority non-runnable job which has been waiting for less time than
bf_min_age_reserve.

Conflicts:
	META
	NEWS

Conflicts:
	src/slurmctld/reservation.c

RESERVE_FLAGS_LIC_ONLY was changed to RESERVE_FLAGS_ANY_NODES

Start job allocation using lowest numbered sockets for block task
distribution for consistency with cyclic distribution.
bug 1540

Continuation of f34c6bbf

block is set up to it's original state.

Add the ability for a compute node to be allocated to multiple jobs, but
restricted to a single user. Added "--exclusive=user" option to salloc,
sbatch and srun commands. Added "owner" field to node record, visible using
the scontrol and sview commands. Added new partition configuration parameter
"ExclusiveUser=yes|no".  Add "Shared=user" job state information.

Bug 1533

Logic was bad in last commit

Fix bug that can permit someone to kill job array belonging to another user.

If a job is submitted with the option --mem=0, then once the job
is allocated resources, reset that value to the amount of memory
on the allocated node with the smallest amount of memory. This is
probably the best option for right now. The downside is that if
the job is requeued, its memory size will no longer be the total
memory on all nodes if the cluster is heterogeneous. On the upside,
some memory data is available for accounting purposes.
bug 1526

Fix support for --mem=0 (all memory of a node) with select/cons_res plugin.
Previously the job would not have it's memory constrained, but it would not
make that memory unavailable to other jobs either.
bug 1526

Conflicts:
	src/sbatch/opt.c

This was previously added to v14.11