Limit the resources used by each CI review
Set their resource requests low, (but try and ensure we don't hit OOMs as a result)
And limit their maxReplicas to 1 more than their minReplicas. (This way we are still testing the pod autoscaling, but we aren't eating review resources)