Testing covariates of Type 2 diabetes-cognition associations in older adults: moderating or mediating effects?