[ROCm] Add support for Punica kernels on AMD GPUs (#3140)

Co-authored-by: miloice <jeffaw99@hotmail.com>

setup.py

@@ -385,12 +385,12 @@ ext_modules = []
 if _is_cuda():
     ext_modules.append(CMakeExtension(name="vllm._moe_C"))
 
-    if _install_punica():
-        ext_modules.append(CMakeExtension(name="vllm._punica_C"))
-
 if not _is_neuron():
     ext_modules.append(CMakeExtension(name="vllm._C"))
 
+    if _install_punica():
+        ext_modules.append(CMakeExtension(name="vllm._punica_C"))
+
 package_data = {
     "vllm": ["py.typed", "model_executor/layers/fused_moe/configs/*.json"]
 }