Add two-tier LLM analysis with triage/escalation

Triage model (LLM_MODEL) handles every message cheaply. If toxicity
>= 0.25, off_topic, or coherence < 0.6, the message is re-analyzed
with the heavy model (LLM_ESCALATION_MODEL). Chat, image analysis,
/bcs-test, and /bcs-scan always use the heavy model.

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

bot.py

@@ -65,12 +65,16 @@ class BCSBot(commands.Bot):
 
         self.config = config
 
-        # LLM client (OpenAI-compatible — works with llama.cpp, Ollama, or OpenAI)
+        # 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)
 
+        # 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)
+
         # Drama tracker
         sentiment = config.get("sentiment", {})
         timeouts = config.get("timeouts", {})
@@ -167,6 +171,7 @@ class BCSBot(commands.Bot):
     async def close(self):
         await self.db.close()
         await self.llm.close()
+        await self.llm_heavy.close()
         await super().close()