reactivate needs to recreate /etc/nologin symlink each time the snapshot is created

[jdreed]

intiscripts apparently scribbled over /etc/nologin (I have no idea how, since this test should not be true: [ ! -L /etc/nologin ] && [ ! -e /etc/nologin ] ). But it did. This caused gdm-config to get an inotify event and now a bunch of cluster machines have "This workstation is out of service". I have pushed gdm-config out, which should gradually fix things, but still.

We should come up with a better solution, or decide that logging in during an update is not a bad thing.