fix: exclude ASG terminating instances from cluster readiness check

[almightychang]

Summary

• Add get_asg_terminating_instance_ids() function to query instances in ASG terminating lifecycle states
• Exclude instances in Terminating:Wait or Terminating:Proceed from readiness check validation

Problem

When LoginNodes are deleted (e.g., setting Count: 0), the cluster update fails because:

1. ASG sets desired capacity to 0, instances enter Terminating:Wait state
2. EC2 still reports these instances as running (lifecycle hook hasn't completed)
3. Readiness check includes these instances and expects new config version
4. Instances will never update (they're being terminated) → check fails repeatedly
5. CloudFormation times out and rolls back

Solution

Query ASG lifecycle states before readiness check and exclude instances that are in:

• Terminating:Wait
• Terminating:Proceed

These instances are being removed and should not be validated against the new config version.

Test Plan

• Deploy cluster with LoginNodes (Count: 2+)
• Update cluster to remove LoginNodes (Count: 0)
• Verify cluster update succeeds without waiting for lifecycle hook timeout
• Verify partial scale-down (Count: 4 → 2) also works correctly

Fixes aws/aws-parallelcluster#7172