this fix the problem with the jobs running up to the end-time cancellation for virgo3 by using a cleaner implementation relying on the SLURM environment variables
virgo2 would still need another method as there those variables are not present
my original method using squeue seems to be broken since last week or this week, maybe after some SPACK related environment variable was removed from the bare-metal submit nodes environment (at least the container does not seem to have changed based on cvmfs filesystem times)
=> This would bring back the weekly test for virgo3 in CDASH but probably at the same time the ones for virgo2 will not be uploaded anymore, so we will not have a week with tests overlap for comparison
=> Still the right direction to go I think, I would then make another MR to cleanup the script of the dirty hack in 3-4 weeks once virgo2 is gone
