Speeding up and reducing memory usage for scientific machine learning via mixed precision