Machine learning-based construction of an oxidative stress-related model reveals molecular subtypes and tumor microenvironment infiltration signatures in colon cancer