Sample from the BART prior #200
Description
Find an easy to sample from the BART prior and document this in a vignette.
The quickest path here is probably fitting BART on a dataset with all zero outcomes and appropriately chosen weights to ensure no "data" is used during split evaluation or leaf parameter sampling. However, right now it is not possible to pass case weights directly through the BART interface, so at minimum this functionality should to be added.
I still want to think further about whether there's a more elegant way to do this.