Using large language models to suggest informative prior distributions in Bayesian regression analysis