Create Epic for parallelization efforts

[bendnorman]

Fully scope the Prefect parallelization Sloan project. Goals and a description can be found in the grant and our project planning document.

This issue should be completed by 8/7/21.

[bendnorman]

It seems like the main goal for this project is to cut down on the time it takes to run the full ETL, integration, and validation tests.

There are still some questions we need to answer before we can scope this epic:

• What existing parts of our pipeline take a long time?
• How should we measure performance?
• Should this first pass include running the pipeline in the cloud? If so, do we want to leverage multiple machines?

[bendnorman]

Mostly covered by #840 and #899.

• What existing parts of our pipeline take a long time? _Trenton did preliminary profiling of the ETL and found that writing PUDL and CEMS to csvs takes foreeever. Ripping out datapackages is being handled in 1176._
• How should we measure performance? This should be further discussed. We have many infrastructure projects going on that will affect performance. It could be nice to include a profiler in our CI.
• Should this first pass include running the pipeline in the cloud? If so, do we want to leverage multiple machines? We decided to focus on getting the Prefect ETL running on one machine first and then move to multiple machines. I believe Jan wrote the PR to handle multiple machines (Full Dask Executor and enabling diskless ETL).