Is there a way in rstanarm to specify priors over model parameters (e.g. model coefficents) that are drawn from different probability distribution families (e.g. Cauchy, Gaussian, Geometric, etc.) rather than simply letting the scale and location parameters vary? While it is straightforward to pass a vector of values to the prior_location and prior_scale argument of the model functions of the package, this assumes that all of the priors over the coefficients are e.g. Gaussian, rather than some being Gaussian and others being Poisson, etc.