[Bug] `deep-health-check-passed-nodes-only` generates incorrect nodeSelector label

[FarhanTejani]

The deep_health_check_passed_nodes_only option generates an incorrect nodeSelector in the Kubernetes job manifest, causing jobs to stay Pending indefinitely.

The Jinja template currently renders:

nodeSelector:
 deep-health-check-passed: "true"

But the actual label on HyperPod EKS nodes is:

nodeSelector:
 sagemaker.amazonaws.com/deep-health-check-status: "Passed"

CLI version: v3.7.0