Fix client not terminating correctly while doing a reboot as part of an update

For unknown reasons, the client does not terminate when it is told to by systemd at reboot time, and one has to wait the entire 1m30s or 3m (depending on setup) interval for it to use kill -9. It doesn't misbehave, but it causes all updates to take much longer.

I saw the issue on the feature-client_auth_api branch, but Manuel Zedel said that he also saw it on master.