Distilling and exploiting quantitative insights from Large Language Models for enhanced Bayesian optimization of chemical reactions