重构: agentic 架构(run_agentic_loop 抽 finish_round_exit + router 调用点 + provider_pool 亲和 enabled)

- run_agentic_loop: 抽 finish_round_exit + emit_ai_completed_once(6 退出点收敛)
- compress/title/knowledge_inject/project: router estimated_context+tier 调用点对齐
- provider_pool: 亲和加 model.enabled 过滤

src-tauri/src/commands/ai/compress.rs

@@ -77,6 +77,7 @@ pub(crate) async fn compress_via_llm(
         modalities: vec![Modality::Text],
         needs_tool_use: false,
         estimated_context: 0,
+        tier: None,
     };
     let model = select_model_id(&compress_req, &provider_config.model_configs)
         .unwrap_or_else(|| provider_config.default_model.clone());