Azure ML training pipeline improvement comments - facilitates better user onboarding for the pipeline

[maxkazmsft]

Address the following comments:

.env.example
Might want to put the min_nodes as 0, so the AML compute cluster will not incur costs after idle period is reached

README.md
Might want to edit it so that it is specific to the Azure VM SKU in which K80/etc maps to and actual VM type (don't just name the GPU type)

interpretation/tests/test_train_pipeline.py
In function _create_config_without make sure name of function seems is complete (right now the name is very vague). Might want to consider renaming it to something more descriptive