Modified all of the models to increase the level of vectorization and thereby increase sampling efficiency.
Also added an argument to prepare_model() that allows the user to decrease the size of the stored model file by removing the count-level predictions. Default is to include the predictions so that posterior and prior predictions still work by default.
Added a scaling factor to the spatial models to ensure consistent interpretation of the priors for the spatial variance across any spatial graph. Following recommendations in https://doi.org/10.1016/j.sste.2018.04.002.
Also improved the internal spatial overlay processes to more accurately assign routes to spatial strata. Done in anticipation of the release of an updated route-path dataset for all US and Canada routes (expected before next annual data release).