Use virtio to avoid VirtualBox issues under load

On an Ubuntu 16.04 host, suring and after load spikes (which come
naturally as part of spinning up kubernetes components, etcd, armada,
etc.), the virtual hardware that VirtualBox provides in the provided
Vagrant example would frequently and randomly become unstable.

The host network that nodes communicate on remained stable, so nodes
could speak to one another and even occasionally join successfully (if
the joining node's public interface hadn't been disrupted so that it
could download images).

The additional load caused by the second etcd cluster for Calico made
this issue occur so often that joining more than one node became quite
unlikely

Switching the VMs to use a paravirtualized network interface seems to
have solved the issue.

I never experienced this issue on Ubuntu 17.04, and it sometimes occurs
without running calico on its own etcd.

Vagrantfile

@@ -16,6 +16,7 @@ EOS
   config.vm.provider "virtualbox" do |vb|
     vb.cpus = 2
     vb.memory = "2048"
+    vb.customize ["modifyvm", :id, "--nictype1", "virtio"]
   end
 
   config.vm.define "n0" do |c|