Text this: Revisiting code readability improvement with LLMs: A critical assessment of fine-tuned models