Support configurable preStop / graceful shutdown for RestateCluster pods

[shamil]

Summary

Support configurable graceful shutdown behavior for RestateCluster pods so users do not need to manually patch the operator-managed StatefulSet.

Problem

After a RestateCluster was provisioned by the operator, I had to manually patch the generated StatefulSet to add a preStop sleep. Without that delay, during a rolling restart the cluster becomes briefly unavailable, which causes errors for Restate services.

Because the StatefulSet is operator-managed, manual patches are not a durable solution.

Current behavior

The operator-generated StatefulSet appears to set a fixed pod-level termination grace period, but it does not expose a way to configure a container lifecycle.preStop hook for Restate pods.

Today the generated pod spec includes termination_grace_period_seconds: Some(60), but no configurable lifecycle.preStop for the Restate container.

Expected behavior

RestateCluster should expose a declarative way to configure graceful shutdown behavior for the managed pods so rolling restarts do not briefly interrupt cluster availability.

Possible approaches:

• Add a simple field such as spec.compute.preStopSleepSeconds
• Or expose a more general field such as spec.compute.lifecycle.preStop
• Optionally also allow configuring spec.compute.terminationGracePeriodSeconds

Additional context

I observed this when the cluster had already been provisioned and then later underwent a rolling restart. Adding a preStop sleep manually to the managed StatefulSet avoided the brief unavailability window.