Textmeddelande: From Bayesian Nonparametrics to Symbols: Scalable and Accurate Model-Free Variable Selection for High-Dimensional Symbolic Regression