Estimating a multivariate normal mean with a bounded signal to noise ratio under scaled squared error loss