A remark on conformal $\SU(p,q)$-holonomy