feat(api): 收敛对话与记忆流程边界，引入 LLM 网关与专用服务

- MemoryService 异步路径委托 MemoryIngestService / MemoryRetrievalService；富化派发经 MemoryEnrichmentScheduler
- WebSocket pipeline 经 ChatTurnService 与显式 DTO 编排单轮对话；回忆录片段入队由 MemoirIngestScheduler 封装
- 新增 LlmGateway（LlmUseCase），各 agent、任务与适配器对齐 ports
- 补充 memory 提示适配、runtime 类型、memory-retrieval 文档、ai-touchpoints 说明与扫描脚本及配套测试

Made-with: Cursor

api/app/ports/llm.py

@@ -12,8 +12,13 @@ class LLMProvider(Protocol):
         *,
         temperature: float = 0.7,
         model: str | None = None,
+        max_tokens: int | None = None,
     ) -> str:
-        """Single-turn completion, returns full response text."""
+        """Single-turn completion, returns full response text.
+
+        ``max_tokens`` when set is passed to the underlying chat API (adapter-specific).
+        """
+
         ...
 
     def stream(
@@ -22,6 +27,8 @@ class LLMProvider(Protocol):
         *,
         temperature: float = 0.7,
         model: str | None = None,
+        max_tokens: int | None = None,
     ) -> AsyncIterator[str]:
         """Streaming completion, yields text chunks (async generator)."""
+
         ...