Add kysrpex.systemd_nspawn role

[kysrpex]

Add kysrpex.systemd_nspawn role, for running containers using systemd-nspawn.

The role is intended be used to run an updated version of HTCondor, taking the roles of Central Manager and Submit.

A separate PR with a playbook that runs the role on the headnode and configures ssh access to the container is also coming.

HTCondor will be installed on the container using Ansible just like it would be done on a regular host.

[bgruening]

This looks in general ok to me. I'm just wondering if we should at first spin up a htcondor 10 cluster and test it. It seems a bit complex in case we decide later to not use it.

[kysrpex]

This looks in general ok to me. I'm just wondering if we should at first spin up a htcondor 10 cluster and test it. It seems a bit complex in case we decide later to not use it.

I just did this with a slightly modified version of #951, on two hosts on our admin OpenStack project (no container, although that is the point of the container, being as transparent as a normal host while permitting the installation of two HTCondor central managers) and jobs can be submitted from the central manager + submit machine to the execute machine, and executed afterwards. I don't know why we would expect it not to work.

I will open a edit the existing meta issue with all details of the plan.