Add paddle to interoperability chapter (NVIDIA#144)

We support `paddle.Tensor.__cuda_array_interface__`(PaddlePaddle/Paddle#68192) and `paddle.to_tensor(tenosr_with_CAI_impl)`(PaddlePaddle/Paddle#69913), so we can convert paddle Tensor from/to other DL frameworks via CAI(based on CAI version 2).

Author: HydrogenSulfate

docs/source/user/cuda_array_interface.rst

@@ -513,6 +513,7 @@ The following Python libraries have adopted the CUDA Array Interface:
 - `ArrayViews <https://github.com/xnd-project/arrayviews>`_
 - `JAX <https://jax.readthedocs.io/en/latest/index.html>`_
 - `PyCUDA <https://documen.tician.de/pycuda/tutorial.html#interoperability-with-other-libraries-using-the-cuda-array-interface>`_
+- `Paddle <https://www.paddlepaddle.org.cn/>`_
 - `DALI: the NVIDIA Data Loading Library <https://github.com/NVIDIA/DALI>`_ :
 
     - `TensorGPU objects