Support 'minReadySeconds' for deployments

[otrosien]

We need to loop back to kubernetes to wait progressing with a rolling restart until the traffic fade-in of a pod is complete. If we don't, we create a skew of the traffic distribution which can cause single pods to overload, in case the rolling-restart is progressing faster than the fade-in.

Ideally, minReadySeconds should automatically be set equivalent to the fade-in time, but if that's not achievable we need to expose minReadySeconds on the manifest for people to set it explicitly.

I know that rolling restart is usually not the way how to operate a stackset but rather deploying a new stack and shifting traffic, but it's also possible to do things like 'kubectl rollout restart deployment'.

Ref https://kubernetes.io/docs/concepts/workloads/controllers/deployment/#min-ready-seconds