Using predicted marginal effects to assess the impact of rurality and free and reduced lunch eligibility on a school-based nutrition intervention