[v0.9.1][DP][V1] Fix rank set in DP scenario & Bump torch-npu version to 2.5.1.post1.dev20250528

[MengqingCao]

What this PR does / why we need it?

Cherry-pick form #1235

1. Fix rank set in DP scenario. The new poc version of torch-npu support setting ASCEND_RT_VISIBLE_DEVICES dynamically, thus we could use the rank set in DPEngineCoreProc directly instead of calculating local rank across dp by hand in the patched _init_data_parallel

Closes: #1170

2. Bump torch-npu version to 2.5.1.post1.dev20250528

Closes: #1242
Closes: #1232

How was this patch tested?

CI passed with new added test.

[Yikun]

just a soft cherrypick

[MengqingCao]

This should merged after #1234

[MengqingCao]

CI failed due to not compatible with V0.9.0, will fix this later

[ganyi1996ppo]

Looks good

[MengqingCao]

@ganyi1996ppo DP will raise timeout error on A2 with this pr, thus we just skip the new added ut currently. And this could fix DP on A3, could you merge this now? The CI except for dp have all passed in https://github.com/vllm-project/vllm-ascend/actions/runs/15700818646/job/44238373529, p.s., tests/multicard/test_dynamic_npugraph_batchsize.py::test_models[True-0.0-64-2-Qwen/Qwen2.5-0.5B-Instruct] failed due to the failure of tests/multicard/test_data_parallel.py::test_data_parallel_correctness[32-Qwen/Qwen2.5-0.5B-Instruct], which dosen't tear down correctly

[ganyi1996ppo]

@ganyi1996ppo DP will raise timeout error on A2 with this pr, thus we just skip the new added ut currently. And this could fix DP on A3, could you merge this now? The CI except for dp have all passed in https://github.com/vllm-project/vllm-ascend/actions/runs/15700818646/job/44238373529, p.s., tests/multicard/test_dynamic_npugraph_batchsize.py::test_models[True-0.0-64-2-Qwen/Qwen2.5-0.5B-Instruct] failed due to the failure of tests/multicard/test_data_parallel.py::test_data_parallel_correctness[32-Qwen/Qwen2.5-0.5B-Instruct], which dosen't tear down correctly

Why a2 failed on the dp case? Dose this failure related to torch_npu?

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[MengqingCao]

Why a2 failed on the dp case? Dose this failure related to torch_npu?

I'll look into this and fix it in next pr. Let's merge this for fixing on A3 firstly

Sorry for the wrong info, there is no bug on A2.

This failure is caused by the wrong method of enabling dp in this pr. And this has been fixed in #1273 cc @ganyi1996ppo