R script to make predictions based on INLA modeling

Obtains predictions from models of Table 2, runs AUC analyses of pixel predictions, and prints predictions of model without predictor (no influence of distance to projects), model using the nearest project as random effect along with influence of distance to projects (pred2), and the model
using municipality intercepts as random effects along with with influence of distance to projects (pred6). The latter is the best-fit model. The script also saves the three-model results as GeoTIFF rasters, the best model is listed as best_auc_hr.tif. NOTE: This R script must run after d2pro3_hires.R, or after loading d_models3.Rdata