Beyond Visual Understanding: Introducing PARROT-360V for Vision Language Model Benchmarking