On the height of the Sylvester resultant