Fix reverse proxying

[avojak]

The current reverse proxy setup doesn't make sense - all requests are forced to go through the single Traefik instance which terminates the SSL, but then the HTTP requests are sent to a different box. Instead, DNS should resolve to the VM hostname, and each VM should have its own reverse proxy. The Traefik playbook should be extensible enough to support this currently (i.e. nothing is specific to the once instance currently in-use).