[Bugfix]Fix deepseek 3.2 C8 precision by rotary tensor (#7537)

### What this PR does / why we need it?
During the attention quantization process of DeepSeek V3.2, it is
necessary to retrieve the Hadamard matrix from the weights to facilitate
the computation.

### Does this PR introduce _any_ user-facing change?
No. But there will be two new tensor in quant weight.

### How was this patch tested?

- vLLM version: v0.18.0
- vLLM main:
8b6325758c

---------

Signed-off-by: mayumeng <m30059191@china.huawei.com>
Co-authored-by: mayumeng <m30059191@china.huawei.com>

vllm_ascend/patch/worker/patch_weight_utils.py

@@ -39,7 +39,16 @@ def patch_deepseek(module):
     def new_remap(name: str, params_dict: dict):
         name = ori_maybe_remap_kv_scale_name(name, params_dict)
 
-        replace_scale_names = ["fa_q.scale", "fa_k.scale", "fa_v.scale", "fa_q.offset", "fa_k.offset", "fa_v.offset"]
+        replace_scale_names = [
+            "fa_q.scale",
+            "fa_k.scale",
+            "fa_v.scale",
+            "fa_q.offset",
+            "fa_k.offset",
+            "fa_v.offset",
+            "indexer.q_rot",
+            "indexer.k_rot",
+        ]
 
         for scale_name in replace_scale_names:
             if name.endswith(scale_name):