On Improved Regret Bounds In Bayesian Optimization with Gaussian Noise