[Bugfix] fix accuracy problem for quantized deepseek models (#768)

### What this PR does / why we need it?

The root cause of the bug is that numerical computations involving NaNs
cannot eliminate them. We addressed it by using `masked_fill_` to
eliminate NaNs while avoiding memory-wasting `torch.where` approach.

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

### How was this patch tested?
This patch was tested with vllm v0.8.5 and vllm-ascend master. I run
deepseek_v3 model with offline inference scripts
(examples/dp_offline/run_dp.sh & data_parallel.py).

Signed-off-by: linfeng-yuan <[email protected]>

Author: linfeng-yuan

vllm_ascend/quantization/w8a8_dynamic.py

@@ -285,7 +285,8 @@ def fused_experts(hidden_states: torch.Tensor,
         valid_token_mask = torch.arange(
             0, sorted_token_indices.shape[0],
             device=device).unsqueeze(1) < num_valid_tokens
-        down_out_list.mul_(valid_token_mask)
+        down_out_list = down_out_list.masked_fill_(~valid_token_mask,
+                                                   0).to(dtype)
         final_hidden_states.index_add_(0, sorted_token_indices, down_out_list)
     else:
         # TODO: Reorder device memory 2 times here, replace the current