check for already running openvpn instance

we did not reimplement this after refactor.

We should:

1. check for an already running openvpn instance (service under osx, differences?)
2. try to kill it (if it was launched by us, we can attach to its management interface and send sigterm...)
3. if we cannot kill it (there is the root special case, in which little we can do, signal unrecoverable error to user)

https://leap.se/code/issues/606

(from redmine: created on 2013-06-18, closed on 2013-07-22, relates #606 (closed))