InternVQA: Advancing Compressed Video Quality Assessment with Distilling Large Foundation Model