Beyond Functional Correctness: Investigating Coding Style Inconsistencies in Large Language Models