Prevent `gitlab-ctl reconfigure` to restart services automagically
This is impacting production right now.
We need to detach the gitlab-ctl reconfigure from a service restart such as redis or postgresql.
We had multiple instances of this, and this has led us to disable chef in the redis and postgres hosts because we are incurring in downtime any time that chef-client gets executed.
- https://gitlab.com/gitlab-com/infrastructure/issues/751
- https://gitlab.com/gitlab-com/infrastructure/issues/1388
- https://gitlab.com/gitlab-com/infrastructure/issues/1517
We need to have omnibus not triggering any more service restarts, so we can own when do we want to ship and apply a new configuration to production.