add: LLM output debug script

test_llm.py

@@ -0,0 +1,16 @@
+import asyncio
+import sys
+sys.path.insert(0, '/workspace/teememo-synth')
+from client import VLLMClient
+
+async def test():
+    async with VLLMClient() as client:
+        result = await client.chat_single(
+            messages=[{'role': 'user', 'content': '日本の中高生の悩みを3件、JSON配列のみで出力してください。'}],
+            max_tokens=1024,
+            enable_thinking=False,
+        )
+        print('RAW OUTPUT:')
+        print(repr(result[:500]))
+
+asyncio.run(test())