Near-Optimal Multiple Testing Procedure In Bayesian Linear Models