【main】SP For Qwen3 MoE (#2209)

### What this PR does / why we need it?
Qwen3 MoE supports SP. In scenarios like AlltoAll, AlltoAllv, and MC2,
replacing AllReduce with Reduce-Scatter and AllGather achieves
computational benefits in norm operations while saving one AllGather
communication. This feature is enabled during the P-phase and delivers
notable gains in long-sequence scenarios (e.g., 16k–25k), with
performance improvements reaching 5%–10%.
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?
``` 
compilation_config={
    "pass_config":{
        "enable_sequence_parallelism": True
    }
},
enable_expert_parallel=True,
```

- vLLM version: v0.10.0
- vLLM main:
9edd1db02b

---------

Signed-off-by: libaokui <libaokui@huawei.com>
Co-authored-by: libaokui <libaokui@huawei.com>

tests/e2e/multicard/test_offline_inference_distributed.py

@@ -234,3 +234,27 @@ def test_models_distributed_DeepSeek_W4A8DYNAMIC():
             },
     ) as vllm_model:
         vllm_model.generate_greedy(prompts, max_tokens)
+
+
+def test_sp_for_qwen3_moe() -> None:
+    example_prompts = [
+        "Hello, my name is",
+    ]
+    sampling_params = SamplingParams(max_tokens=5,
+                                     temperature=0.0,
+                                     top_k=50,
+                                     top_p=0.9)
+
+    with VllmRunner(
+            snapshot_download("Qwen/Qwen3-30B-A3B"),
+            dtype="auto",
+            tensor_parallel_size=2,
+            distributed_executor_backend="mp",
+            compilation_config={
+                "pass_config": {
+                    "enable_sequence_parallelism": True
+                }
+            },
+            enable_expert_parallel=True,
+    ) as vllm_model:
+        vllm_model.generate(example_prompts, sampling_params)