Learning from Failure: Predicting Electronic Structure Calculation Outcomes with Machine Learning Models