Create a tensor adapter for mlx arrays in src/onnx_ir/tensor_adapters.py. Implement all the methods and follow existing examples. Create proper tests.