feat(memory): add LLM-driven memory extraction from conversations

- memory/extractor: async LLM-based memory extraction from chat history
- Parses JSON lines from LLM response as memory entries
- Wire into orchestrator pipeline via PostProcess hook
- Non-blocking: runs in goroutine with 30s timeout
- Automatic dedup via existing keyword overlap logic

backend/cmd/cyrene/main.go

@@ -182,6 +182,11 @@ func main() {
 	// === 11. AI-Core: Orchestrator ===
 	semanticCache := llm.NewSemanticCache()
 	orch := orchestrator.NewPipeline(llmSelector, semanticCache, ctxStore, subMgr, personaInjector)
+	if memSvc != nil {
+		memAdapter := llmSelector.DefaultAdapter()
+		memExtractor := memory.NewExtractor(memSvc, &memory.LLMAdapterBridge{Adapter: memAdapter})
+		orch.SetMemoryExtractor(memExtractor)
+	}
 	log.Printf("[orchestrator] pipeline ready")
 
 	// === 11.5 Platform Bridge ===