Fix stress parallel deployment deadlock issue

When using the parallel feature (indexed jobs) in stress, we gate ARM deploys on pod 0 completing deployment to avoid race conditions. There is a deadlock issue where many pods are deployed exhausting our node resource quota. If pod 0 does not get scheduled, then pods 1+n will use up all quota polling for pod 0 completion, but pod 0 will never be scheduled.

Locate the stress parallel indexed-job deployment code that gates ARM deploys on pod 0, then inspect how pods poll for completion and consume node quota. Reproduce the case where pod 0 is unscheduled; done means the remaining pods cannot exhaust the quota while waiting and the deployment can recover or terminate clearly.

Written by the indexing model from the issue text.