How Hard is Two-Level Logic Minimization: an Addendum to Garey & Johnson