lightly improves security of passwords and user registration/login
removes num included/excluded tracking on review records, and instead computes those values on-the-fly as needed; this allows us to remove a lot of bookkeeping logic, and improve overall performance while reviewing studies
orders write-only collection relationships for deterministic results; fixes a bug in user related reviews; they're now returned as a not-smart list
slightly improves study citation text content generation, in both python and sql contexts
fixes notable bug updating studies' number of citation/fulltext reviewers after review update, where fulltext updates were consistently overwriting citation updates when we actually needed both
adds lots of unit tests for core db models (user, review, study), which were only implicitly tested through api results
changes