A More Flexible BART Model // Implements a faster and more expressive version of Bayesian Additive Regression Trees that, at a high level, approximates unknown functions as a weighted sum of binary regression tree ensembles. Supports fitting (generalized) linear varying coefficient models that posits a linear relationship between the inverse link and some covariates but allows that relationship to change as a function of other covariates. Additionally supports fitting heteroscedastic BART models, in which both the mean and log- variance are approximated with separate regression tree ensembles. A formula interface allows for different splitting variables to be used in each ensemble. For more details see Deshpande (2025) doi:10.1080/10618600.2024.2431072 and Deshpande et al. (2024) doi:10.1214/24-BA1470.