feat: try out cpu vllm

Dockerfile

@@ -3,7 +3,10 @@ FROM ghcr.io/ten-framework/ten_agent_build:0.4.17-2-geca93b3-cuda
 RUN pip install hf_transfer
 
 # install vllm
-RUN pip install vllm
+#RUN pip install vllm
+RUN VLLM_TARGET_DEVICE=cpu 
+    PIP_EXTRA_INDEX_URL=https://download.pytorch.org/whl/cpu \
+    pip install https://github.com/vllm-project/vllm
 
 # install node and npm for demo
 ENV NVM_DIR /usr/local/nvm