Variations in sociodemographic features, exposure characteristics and genetic polymorphisms between community and hospital controls: a validation study of control selection in case-control designs