Adding AutoML in parallel to HyperDrive

[lenisha]

• [x] add pipeline step for model sweep AutoML

[hosseinsarshar]

• The train_trainsformer.py (hyperdrive) script evaluating the model based on F1 score, however, the AutoML currently doesn't support this metric. So a quick fix is to change the hyperdrive evaluation method to AUC.
• No evaluation on test set is possible at this time, there needs to be subsequent step which evaluates the model against the test set.

[lenisha]

PR for updated Automl + Test step PR

More feedback items:

• AutoML generated Model conda_env.yml fails installing pytorch framework
• Need documentation on registering and deploying AutoML generated model (score.py, conda env, etc) details