core: implement Float::to_precision in C without callbacks