[CI/Build] Investigate torchrun distributed tests hanging issue (#33650)

Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>

tests/distributed/test_torchrun_example_moe.py

@@ -39,6 +39,9 @@ llm = LLM(
     gpu_memory_utilization=random.uniform(0.7, 0.9),
     swap_space=random.randint(1, 4),
     seed=0,
+    # FIXME(Isotr0py): async scheduling causes deadlock
+    # on torchrun with PP, need to investigate further.
+    async_scheduling=False,
 )
 
 outputs = llm.generate(prompts, sampling_params)