merge versions of prd_pipeline into a single accessible folder

[stevehadd]

the idea with that script is there should be only one version that all notebooks refer to (rather than multiple copies) , so we need to think about how we do that so that it works with azureml clusters.