feat: add KEDA/HPA managed scaling support

[m1rl0k]

Added opt-in KEDA/HPA scaling support via temporal.io/keda-managed annotation on TemporalWorkerDeployment resources.
When this annotation is present:

• Non-zero scale operations are filtered out (only scale-to-zero is preserved for sunset behavior)
• Existing replicas are preserved during deployment updates to avoid fighting KEDA
  This is a ~17 line change in internal/controller/genplan.go that is fully backwards compatible - existing users without the annotation see zero behavior change.

Why?
Users running Temporal workers on Kubernetes often want to use KEDA for autoscaling based on Temporal task queue depth metrics. Currently, the operator's replica management conflicts with KEDA's scaling decisions, causing a "fight" between the two controllers.

This change allows the operator to defer scaling responsibility to KEDA while still handling:

• Version management and rollouts
• Scale-to-zero for deprecated/sunset versions
• All other non-scaling lifecycle operations

Checklist

1. Closes #N/A (new feature request)

2. How was this tested:

   • Code review of change paths
   • Verified annotation check is opt-in (no impact to existing users)
   • Logic review: only affects ScaleDeployments and UpdateDeployments replica handling when annotation present

3. Any docs updates needed?

   • README update recommended to document temporal.io/keda-managed annotation
   • Example KEDA ScaledObject configuration for Temporal task queues would be helpful

[CLAassistant]

All committers have signed the CLA.

[m1rl0k]

@jlegrone any objections?