[Nightly][Test] Add Qwen3-Next-80B-A3B-Instruct-W8A8 nightly test (#5616)

### What this PR does / why we need it?
There was an accuracy issue with **Qwen3-Next-80B-A3B-Instruct-W8A8**
model in the old version of **Triton-Ascend**, so, we are now adding one
nightly test to maintain it.
### Does this PR introduce _any_ user-facing change?
N/A
### How was this patch tested?

- vLLM version: v0.13.0
- vLLM main:
7157596103

Signed-off-by: IncSec <1790766300@qq.com>

.github/workflows/nightly_test_a3.yaml

@@ -132,6 +132,9 @@ jobs:
           - name: qwen3-235b-w8a8
             os: linux-aarch64-a3-16
             tests: tests/e2e/nightly/single_node/models/test_qwen3_235b_w8a8.py
+          - name: qwen3-next-w8a8
+            os: linux-aarch64-a3-4
+            tests: tests/e2e/nightly/single_node/models/test_qwen3_next_w8a8.py
           # TODO: Replace deepseek3.2-exp with deepseek3.2 after nightly tests pass
           # - name: deepseek3_2-exp-w8a8
           #   os: linux-aarch64-a3-16