Evidence for an upper limit to mitotic spindle length