Bayesian Bandit Algorithms with Approximate Inference in Stochastic Linear Bandits