[main][bugfix] bugfix for qwen3 moe quantization (#4599)

### What this PR does / why we need it?
Fix the issue where the qwen3 moe service cannot be started due to
upgrading the vllm version

Error info:
AttributeError: 'AscendFusedMoE' object has no attribute 'use dp
chunking'

### Does this PR introduce _any_ user-facing change?
no


- vLLM version: v0.11.2

---------

Signed-off-by: Wang Kunpeng <1289706727@qq.com>

vllm_ascend/quantization/quant_config.py

@@ -408,11 +408,10 @@ class AscendFusedMoEMethod(FusedMoEMethodBase):
         quant_config: The Ascend quantization config.
     """
 
-    def __init__(self,
-                 quant_config: AscendQuantConfig,
-                 prefix: str,
-                 packed_modules_mapping: Dict[str, Any],
-                 layer: torch.nn.Module = None):
+    def __init__(self, quant_config: AscendQuantConfig, prefix: str,
+                 packed_modules_mapping: Dict[str,
+                                              Any], layer: torch.nn.Module):
+        super().__init__(layer.moe_config)
         self.quant_method = get_quant_method(quant_config.quant_description,
                                              prefix,
                                              "moe",