How visualizing inferential uncertainty can mislead readers about treatment effects in scientific results