Efficient Bayesian Sampling Using Normalizing Flows to Assist Markov Chain Monte Carlo Methods