Support Pangu Pro MoE model

[Angazenn]

What this PR does / why we need it?

Support Pangu Pro MoE model (https://arxiv.org/abs/2505.21411)

Does this PR introduce any user-facing change?

Yes, new model supported

How was this patch tested?

Test locally

[shen-shanshan]

@Angazenn Thanks for your contribution, please also update the model support doc: https://github.com/vllm-project/vllm-ascend/blob/main/docs/source/user_guide/supported_models.md.

[shen-shanshan]

@Angazenn Please modify the model name:

PanGuMoEModel -> PanguProMoEModel
PanGuMoEForCausalLM -> PanguProMoEForCausalLM

The name of opensource config is different.

[Yikun]

VLLM_USE_V1=1 vllm serve /root/.cache/pangu-pro-moe-model \
    --tensor-parallel-size 4 \
    --swap-space 16 \
    --disable-log-stats  --disable-log-requests  \
    --trust-remote-code  --enforce-eager

curl http://localhost:8000/v1/completions \
    -H "Content-Type: application/json" \
    -d '{
        "model": "/root/.cache/pangu-pro-moe-model",
        "prompt": "The future of AI is",
        "max_tokens": 128,
        "temperature": 0
    }'

I do an E2E test in my local env, it works as expected.

cc @ganyi1996ppo @wangxiyuan