Dual-stack case is broken for Vagrant-based test cluster

[antoninbas]

Describe the bug
Trying to run ./provision -ip-family dual and deploying Antrea (v1.4) in the created cluster will not work. Antrea Agents will keep crashing because of an initialization error:

I1221 21:17:14.361053       1 log_file.go:99] Set log file max size to 104857600
I1221 21:17:14.361914       1 agent.go:72] Starting Antrea agent (version v1.5.0-dev-aa018fa8.dirty)
I1221 21:17:14.361961       1 client.go:79] No kubeconfig file was specified. Falling back to in-cluster config
I1221 21:17:14.363310       1 client.go:79] No kubeconfig file was specified. Falling back to in-cluster config
I1221 21:17:14.363761       1 prometheus.go:171] Initializing prometheus metrics
I1221 21:17:14.364282       1 ovs_client.go:67] Connecting to OVSDB at address /var/run/openvswitch/db.sock
I1221 21:17:14.365134       1 agent.go:338] Setting up node network
F1221 21:17:14.377570       1 main.go:58] Error running agent: error initializing agent: failed to get local IPNet device with IP &{192.168.77.100 fd3b:fcf5:3e92:d732::100}: IPs of localIPs should be on the same device

This was because #2436 introduced the limitation that both IP addresses have to be assigned to the same device on the Node. This was deemed a reasonable limitation during review. However, for the Vagrant-based test cluster, this condition is not meant: 2 different networks are created for the different address families.

To satisfy the condition, we need to find a mechanism to assign both IPs to the same device. It's unlikely that there is a standard way to do this in the Vagrantfile, we may have to run a script in each Vagrant VM to move the IP.