Hosts running builder VMs moved to an internal subnet.

As we've agreed to not touch the storage for now, at least hosts using it should remain in the external subnet. All the Jenkins hosts however can be moved into the infra VLAN.

6 hypervisors already moved and running Jenkins workers.

As part of this ticket I installed a fresh HE on iSCSI on a cluster consisting of ovirt-srv11 and ovirt-srv12

My notes are as follows:
1) it is not trivial to set up HE storage on an isolated non-routable VLAN.
I had to manually add a VLAN subinterface on both hosts, setup HE, add second host, and then from the UI I was able to set up host networks on the idle one. However for repeating the process I had to move HE between hosts and hit a bug similar to bz#1362618. The host got stuck in "preparing for maintenance" state and the VM was reported running on both hosts by VDSM although it was migrated successfully and the source host had no active qemu processes on it. I still have problems with host maintenance in that environment (1362618)
2) 4.0 removes the VLAN limitation present in 3.5 so we can use both tagged and untagged ones for VM networks (the initial reason for this ticket)
3) in the current state we would need a working router VM to make oVirt work in the internal subnet

As a result I've opened to upgrade the existing engine instead as we're moving builders to the internal subnet and aren't confined in IP address space any more.

We have two limiting factors that limit decisions here:
1) networking team can't do routing for us at the moment -> we need to route between VLANs ourselves
2) oVirt doesn't support mixing tagged and untagged VLAN networks -> the native VLAN can't be routed from within oVIrt

The solution here is to do routing outside of oVirt, however we don't have hardware for that at the moment. For now we'll use a libvirt VM outside of oVirt to do the routing and will work on getting dedicated hardware to do it on.