A bayesian approach to the estimation of the largest normal mean