fix: make "Pod was rejected:" errors transient

[tooptoop4]

Fixes #12572

[tczhao]

This error happens when

• a node in your Kubernetes cluster runs out of disk space
• timing/logic between scheduler and kubelet, usage by other pods in the same node could cause issue when the pod got scheduled but failed to start

I don't believe this is a transient pattern as the cluster node could be struggling for real reason.
Many other error messages could match the "Pod was rejected" pattern, and should be handled properly at the kubelet end.
Argo wf user can use retry: Always to retry such pod

[tooptoop4]

i don't want to retry for every type of error