AdaNCA: Neural Cellular Automata as Adaptors for More Robust Vision Transformer