Gravar-mail: Evaluating the potential role of pleiotropy in Mendelian randomization studies