Describir: Machine learning and feature engineering for predicting pulse presence during chest compressions