[Feature] optimize sp & qwen3 next support sp. (#3225)

This PR will accomplish the following tasks: 
**optimize SP**
In the old version implementation, the first layer was all_reduce, which
used rms to split chunks. We changed it to perform reduce_scatter on the
embedding side, replace one all_reduce operation and one chunk with one
reduce_scatter operation.
**Support qwen3 next**
Since Qwen3 Next includes a linear attention module, the prefix name of
this module cannot take effect directly.


- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: weijinqian_v1 <weijinqian@huawei.com>
Co-authored-by: weijinqian_v1 <weijinqian@huawei.com>

tests/ut/torchair/models/test_torchair_deepseek_v2.py

@@ -276,10 +276,7 @@ def test_torchair_deepseek_v2_mla_attention(mock_rms_norm, mock_distributed,
 @patch("torch_npu.npu_add_rms_norm")
 @patch("torch_npu.npu_rms_norm")
 @patch("torch.ops.vllm.maybe_wait_prefetch_done", side_effect=lambda x: None)
-@patch("torch.ops.vllm.maybe_chunk_residual",
-       side_effect=lambda x, residual: residual)
-def test_torchair_deepseek_v2_decoder_layer(mock_maybe_chunk_residual,
-                                            mock_maybe_wait_prefetch_done,
+def test_torchair_deepseek_v2_decoder_layer(mock_maybe_wait_prefetch_done,
                                             mock_rms_norm, mock_add_norm,
                                             mock_distributed, base_config,
                                             vllm_config, mock_forward_context):