Will They Like This? Evaluating Code Contributions with Language Models