Only set memory.memsw.limit_in_bytes to the computed amount of allowed
RAM+Swap when ConstrainSwapSpace=yes in cgroup.conf.
When ConstrainSwapSpace=yes and ConstrainRAMSpace=no, automatically
set AllowedRAMSpace to 100% in order to compute the memsw.limit based
on the allocated memory plus the allowed swap percent. Then use that
limit for both mem and mem+sw limits.

stop the current poll asap.

having the realtime server go down over and over again while waiting
for the poll to finish.

to the functions instead of the global rt_running flag

host names. Like done elsewhere in SLURM

users/accounts/flags

correctly for current job.

The uptime is extra information not used in the slurm.conf file

realtime server.

correct prefix

start in the future but is able to run now.

Previously the SIGCONT would be generated easily as part of gang
scheduling and cause the job step initiaition to abort.
This is being back-ported from SLURM v2.5/master.

One of our testers discovered an inconsistency between trigger names that
are displayed with 'strigger --get' and the trigger names passed to trigger
programs.

accounting as well.