[feature] Support W8A8 PD-Mix Quantization (#4235)

In PD-separated deployment scenarios:

* MoE layers use dynamic quantization exclusively.
* For the Attention module, Prefill (P) nodes use **dynamic**
quantization, while Decode (D) nodes use **static** quantization.

In PD-mixed deployment scenarios:
* **All components fall back to dynamic quantization**, as it is
difficult to distinguish between Prefill and Decode tokens.
___

- vLLM version: v0.11.2
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.2

---------

Signed-off-by: SlightwindSec <slightwindsec@gmail.com>
Signed-off-by: Slightwind <slightwindsec@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

vllm_ascend/quantization/w8a8_dynamic.py

@@ -60,6 +60,7 @@ class AscendW8A8DynamicLinearMethod:
         params_dict["weight_offset"] = torch.empty(output_size,
                                                    1,
                                                    dtype=params_dtype)
+        params_dict["bias"] = torch.zeros(output_size, dtype=torch.float32)
         return params_dict
 
     def get_pergroup_param(self,
@@ -110,6 +111,7 @@ class AscendW8A8DynamicLinearMethod:
         layer.weight_scale.data = layer.weight_scale.data.flatten()
         layer.weight_scale_fp32 = layer.weight_scale.data.to(torch.float32)
         layer.weight_offset.data = layer.weight_offset.data.flatten()
+        layer.bias.data = layer.bias.data.to(layer.weight_scale.data.dtype)
 
 
 class AscendW8A8DynamicFusedMoEMethod: