VLA-Arena: An Open-Source Framework for Benchmarking Vision-Language-Action Models