Add LLM request/response logging to database

Log every LLM call (analysis, chat, image, raw_analyze) to a new
LlmLog table with request type, model, token counts, duration,
success/failure, and truncated request/response payloads. Enables
debugging prompt issues and tracking usage.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

bot.py

@@ -65,15 +65,18 @@ class BCSBot(commands.Bot):
 
         self.config = config
 
+        # Database (initialized async in setup_hook)
+        self.db = Database()
+
         # LLM clients (OpenAI-compatible — works with llama.cpp, Ollama, or OpenAI)
         llm_base_url = os.getenv("LLM_BASE_URL", "http://athena.lan:11434")
         llm_model = os.getenv("LLM_MODEL", "Qwen3-VL-32B-Thinking-Q8_0")
         llm_api_key = os.getenv("LLM_API_KEY", "not-needed")
-        self.llm = LLMClient(llm_base_url, llm_model, llm_api_key)
+        self.llm = LLMClient(llm_base_url, llm_model, llm_api_key, db=self.db)
 
         # Heavy/escalation model for re-analysis, chat, and manual commands
         llm_heavy_model = os.getenv("LLM_ESCALATION_MODEL", llm_model)
-        self.llm_heavy = LLMClient(llm_base_url, llm_heavy_model, llm_api_key)
+        self.llm_heavy = LLMClient(llm_base_url, llm_heavy_model, llm_api_key, db=self.db)
 
         # Drama tracker
         sentiment = config.get("sentiment", {})
@@ -84,9 +87,6 @@ class BCSBot(commands.Bot):
             offense_reset_minutes=timeouts.get("offense_reset_minutes", 120),
         )
 
-        # Database (initialized async in setup_hook)
-        self.db = Database()
-
     async def setup_hook(self):
         # Initialize database and hydrate DramaTracker
         db_ok = await self.db.init()