avoid duplicate epilog execution

this will happen if a job releases its resources to expand another job,
triggering a second epilog and slurmctld errors about cpu count underflows