Adding performance metrics to AG Assistant

[rob7112]

AGA does not return any performance metrics after training/testing at this point in time. It only returns the predicted labels of the test set, which means that the user needs to compute the metrics separately (metrics such as accuracy, precision, recall for classification and mse, mae for regression).

It would be very helpful to provide those metrics too as part of the output. Specifically, my recommendation is to return a CSV file with the training/validation/testing metrics on a leaderboard, similarly to what AG tabular does for TabularPredictor.leaderboard(), but expanded with auxiliary metrics (accuracy, precision, recall, etc.) like those pulled with TabularPredictor.evaluate(auxiliary_metrics=...).