No changes in logic

The plugin-specific errno handling these were intended for does
not exist, and these ranges can be reused without issue at this point.

No plugins implement this interface, so discard.

Never called. None of the plugins have bothered to implement this correctly.
The jobcomp/script plugin comes closest, but did not pick error codes within
the designated range.

Remove "cons_tres: " from some src/common/gres.c log messages.
They are vestigial from relocated functions.

This needs to be replaced by a finer-grained access control in the
next pre-release of v19.05

Move tests into relevant FastSchedule groups.
Rename tests based on FastSchedule setting.
Tweak some expected output to match actual.
Add capability for expected errors.
Comment out some tests that still need work.

Bug 6635

Bug 6635

Triggered with GresTypes=gpu,mps. zero gpu count for node defined
in slurm.conf and non-zero gpu count in gres.conf

As libssh2 is gone, we no longer need x_ac_ssh2.m4.

Add in x_ac_x11.m4 to restore the --disable-x11 option.

Unconditionally build x11_forwarding.c now that it's no longer dependent
on the libssh2 libraries being available.

Bug 3647.

Leave the state file name - burst_buffer_cray_state alone.

Bug 6440.

The order can change as long as it is kept in sync with the ops struct,
which already notes this requirement.

Reverts code added in 8486c97e to handle failure mode of BGQ jobs.

Bug 6566.

The int size may not be enough when converting from seconds to usec for the
maximums defined for bf_interval. _yield_sleep has been to adapt to this change
too.

Bug 6401

All numeric backfill parameters have now maximums and minimum values.

Bug 6401

Improve log and comments

Set CUDA_VISIBLE_DEVICES and CUDA_MPS_ACTIVE_THREAD_PERCENTAGE environment
variables in Prolog and Epilog for jobs requesting gres/mps. This can be
used to start, terminate and otherwise manage mps servers for the user.

Previous logic would drain node with less than configured gres.
bug 6635

This also removes the check for duplicates that seemed redundant.

Bug 6635

Without this change a single job could not use 100% of the gres/mps on
any single device, but only up to one less than the total count.

bug 6635

Without this change, test19.18 would generate fatal errors due to
using the Slurm APIs, but not having the same environment as the
slurmctld

as the parameter to determine how to gather gres/gpu configuration.
bug 6635

Commit 6ac4ce84 introduces a dependency in the gres plugin
on the select plugin.

Bug 6638.

slurmstepd needed some changes to avoid creating a size zero bitmap.
bug 6635

If we do have 0 this means we have no gres on the node.

Any other call to the bitstring will fail (bit_test, bit_clear, etc).

I was unable to find a situation where this would be valid.

Bug 6616