MI300X_GPTOSS_DEBUG

Issue Reproduction:

1. Launch a container with the image: rocm/vllm-dev:open-mi300-08052025
2. Serve VLLM with VLLM_ROCM_USE_AITER=1 VLLM_USE_AITER_UNIFIED_ATTENTION=1 VLLM_ROCM_USE_AITER_MHA=0 vllm serve openai/gpt-oss-120b --compilation-config '{"full_cuda_graph": true}' &
3. Send prompt request 1st time, output is fine
4. Send same prompt request again, output is never ending "!!!!!!!!!!"

Steps for suggested fix on AITER kernel level:

1. After launching the container, handle aiter change with sh aiter_fix.sh
2. Run a one-line hotfix (sorryy) for import as such: sed -i 's|from aiter.ops.triton.unified_attention import unified_attention|from aiter.ops.triton.hybrid_attention_integration import enhanced_unified_attention as unified_attention|g' /app/vllm-os-mini/vllm/v1/attention/backends/triton_attn.py
3. Try serving and sending requests again with this command: VLLM_FORCE_HYBRID_ATTENTION=1 VLLM_ROCM_USE_AITER=1 VLLM_USE_AITER_UNIFIED_ATTENTION=1 VLLM_ROCM_USE_AITER_MHA=0 vllm serve openai/gpt-oss-120b --compilation-config '{"full_cuda_graph": true}' &