A semiparametric approach for meta‐analysis of diagnostic accuracy studies with multiple cut‐offs