Exploring comparative effect heterogeneity with instrumental variables: prehospital intubation and mortality