Evaluate use of MLflow for experiment tracking

[stevehadd]

Currently we have quite a manual implementation of keeping track of experiements. There are now framework to facilitate this. One of which we have already started using in the form of the AzureML API, but this is AzureML specific. Other frameoworks are more general and it would be to consider rather using one of them, for example MLflow:

• https://mlflow.org/
• https://databricks.com/product/managed-mlflow