refactor: 重构 inference 模块架构，引入设计模式并分组文件

- 新增 protocol.py 协议层，Template Method 模式消除流/非流分支 45% 重复
- SSEBuilder 统一 SSE 构造，StopChecker 独立 stop_sequence 检测
- AnthropicHandler 追踪已产出文本，修复 stop 时重复 delta
- server.py 路由从约 100 行缩减至 3 行
- 拆分为 core/（cache/executor/scheduler/task）和 api/（protocol/server）
- 外部保持二级导入路径（from astrai.inference import Name）
- 删除所有分隔线注释，代码按语义自然分组

astrai/inference/api/__init__.py

@@ -0,0 +1,33 @@
+"""Inference API: protocol handlers and FastAPI server."""
+
+from astrai.inference.api.protocol import (
+    AnthropicHandler,
+    OpenAIHandler,
+    ProtocolHandler,
+    SSEBuilder,
+    StopChecker,
+    StreamContext,
+)
+from astrai.inference.api.server import (
+    AnthropicMessage,
+    ChatCompletionRequest,
+    ChatMessage,
+    MessagesRequest,
+    app,
+    run_server,
+)
+
+__all__ = [
+    "AnthropicHandler",
+    "OpenAIHandler",
+    "ProtocolHandler",
+    "SSEBuilder",
+    "StopChecker",
+    "StreamContext",
+    "AnthropicMessage",
+    "ChatCompletionRequest",
+    "ChatMessage",
+    "MessagesRequest",
+    "app",
+    "run_server",
+]