Health checks and update RollingUpdateStrategy for registry

We deployed the first registry configuration change while taking live traffic production#1127 (closed) and noticed that we saw a pretty big, but brief error spike during the apply.

Observing the apply in realtime the pods serving production traffic were terminated very aggressively.

RollingUpdateStrategy:  25% max unavailable, 25% max surge

Proposed changes to lower errors during the rolling update

• Add readiness/liveness probes to the helm chart gitlab-org/charts/gitlab#1571 (closed) implemented in gitlab-org/charts/gitlab!932 (merged)
• reduce the max unavailable to something much lower, I think we should set this to the absolute number of 1 or lower the percentage
• Allow draintimeout https://github.com/docker/distribution/blob/release/2.7/configuration/configuration.go#L90 to be set in the chart and increase it. Draintimeout added in gitlab-org/charts/gitlab!934 (merged)
• Configure terminationGracePeriod if necessary (this is set to 30 seconds by default) gitlab-org/charts/gitlab!935 (closed)