Galbot's integrated brain-cerebellum-neural-control AI stack, combining three Vision-Language-Action models — GraspVLA for object manipulation, GroceryVLA for r