Optimum Experimental Design for a Bayes Estimator in Linear Regression