Add a script of how to setup in China network

Basically something like this:
https://github.com/vllm-project/vllm/blob/main/use_existing_torch.py

We need do things like:
1. downlaod assets (maybe test assets like llama if possible)
2. set env vars (like those uv ones if there’s a known mirror)
3. set .gitmodules to use github mirror
4. support swanlab: https://github.com/NVIDIA-NeMo/RL/pull/923
5. etc.