Fix broken DNS in arcus CI

[sjpb]

Updates the image used on arcus to one without /etc/resolv.conf to fix broken DNS when booting the image on arcus's ilab-60 subnet. See stackhpc/slurm_image_builder#6 (comment) for the full horrible details.

NB: smslabs should also be updated to use this, but need s3 quota to be able to transfer it.

Note the addition of the connection reset does not appear to be strictly needed, but the above DNS problem caused:

TASK [firewalld : Install firewalld package] ***********************************
task path: /home/runner/work/ansible-slurm-appliance/ansible-slurm-appliance/ansible/roles/firewalld/tasks/install.yml:1
fatal: [ci2624804230-login-0]: UNREACHABLE! => {
    "changed": false,
    "unreachable": true
}

MSG:

Data could not be sent to remote host "10.60.102.246". Make sure this host can be reached over ssh: Could not chdir to home directory /home/rocky: No such file or directory

so although that wasn't the /root/ cause, it seems sensible to try to prevent that.