[Bugfix] Fix Qwen3-Omni MoE backend selection (#7019)

Signed-off-by: hsliuustc0106 <liuhongsheng4@huawei.com>
Co-authored-by: hsliuustc0106 <liuhongsheng4@huawei.com>

vllm_omni/deploy/qwen3_omni_moe.yaml

@@ -27,6 +27,7 @@ stages:
     max_num_batched_tokens: 32768
     max_num_seqs: 64
     gpu_memory_utilization: 0.9
+    moe_backend: triton
     trust_remote_code: true
     enable_prefix_caching: false
     devices: "0"
@@ -38,6 +39,7 @@ stages:
     max_num_batched_tokens: 32768
     max_num_seqs: 64
     gpu_memory_utilization: 0.6
+    moe_backend: triton
     trust_remote_code: true
     enable_prefix_caching: false
     devices: "1"