Removing phenotypic distribution assumptions from tests of linkage disequilibrium for quantitative traits