Update single-node zero downtime instructions to caveat some downtime
Currently our zero-downtime for single-node seem to indicate it is possible: https://docs.gitlab.com/omnibus/update/README.html#single-deployment. As far as I know while we could provide this for some upgrades, we could not for all.
We should update the page, to indicate that they are not possible in particular after Puma. If we can, it would also be good to indicate an expectation for downtime when Puma restarts.
Edited by Joshua Lambert