Fix worker_machineset_suffix default to prevent validation failure

[stencell]

Summary

• [scale_ocp_workers] Set worker_machineset_suffix default from "" to "compute"
• Fixes validation failure when creating new MachineSets with a different instance type

Details

When a user orders a cluster with a worker_instance_type that differs from the pool's existing instance type, the scale_ocp_workers role takes the "create new MachineSets" path. This path validates that worker_machineset_suffix is not empty — but the default was "", causing the task to always fail:

TASK [agnosticd.cloud_provider_aws.scale_ocp_workers : Validate worker_machineset_suffix is provided]
fatal: [localhost]: FAILED! => {"msg": "worker_machineset_suffix is required when creating MachineSets with a different instance type"}

Setting the default to "compute" provides a sensible generic suffix (e.g., ocp-xjgsc-worker-us-east-2a-compute) while still allowing callers to override it for specific use cases like "gpu" or "highmem".

Test plan

• Deploy ocp-cluster-aws with worker_instance_count_param: 2 and a worker_instance_type that differs from the pool's default
• Verify the scale role creates new MachineSets with the -compute suffix
• Verify worker nodes become Ready