StarVLA: A Lego-like Codebase for Vision-Language-Action Model Developing - View it on GitHub