improve scaling of data prep with dask

[stevehadd]

creating a cluster might be helpful, for doing ML as well on azure ML https://cloudprovider.dask.org/en/latest/azure.html

[arh89]

Agreed to put into backlog and re-evaluate whether necessary after @DPeterK's work with dagster

[stevehadd]

our thinking on how to do the data processing has probably movd on now, so would rather close this and pick up the data workflow stuff some other way