Make folds.svy() and folds.svydesign() directly accessible and documented, so that users can form survey-design-based CV folds and use them with other models besides svyglm. Add examples of using them to tune design-consistent random forests from rpms package.
Also: Allow logistic regression response variable to be a factor, not just 0/1 (thanks to Aja Sutton and Michelle Jamieson). Add na.rm arguments, passed internally to svymean when calculating average of CV test losses. Update docs with link to Stat article doi:10.1002/sta4.454, published online on Jan 12, 2022.