Support for deterministic worker node deletion (scale-down) in ROSA HCP clusters

Issue

When scaling down a Machine Pool in a ROSA HCP cluster (e.g., reducing the replica count via CLI or OCM), the underlying system selects a worker node for removal non-deterministically (randomly).

Users lack a supported mechanism to:

• Target a specific node for deletion (e.g., a node used for temporary testing).
• Ensure critical or stateful workloads are not disrupted by random node termination.
• Perform a graceful, manual scale-down where the replica count and the specific EC2 instance stay in sync.