refactor: 重构推理引擎控制逻辑，修复连续批处理核心缺陷

- 修复 decode 阶段新任务覆盖已有任务的严重缺陷
- 修复线程安全问题（热路径无锁竞争）
- 修复前缀缓存引用计数管理不当导致缓存被驱逐
- 修复 pad_id 缺失导致全量 prefill 崩溃
- 修复 RoPE 位置错乱（不同位置任务共用 start_pos）
- 新增 slot 版本追踪实现前缀缓存零拷贝复用
- 新增异步流式生成接口避免阻塞事件循环
- 添加完整英文文档字符串

tests/inference/test_server.py

@@ -100,13 +100,12 @@ def test_chat_completions_non_stream(client, loaded_model, mock_engine, monkeypa
 def test_chat_completions_stream(client, loaded_model, mock_engine, monkeypatch):
     """POST /v1/chat/completions with stream=true returns SSE stream."""
 
-    # Simulate a streaming generator that yields cumulative responses
-    def stream_gen():
+    async def async_gen():
         yield "cumulative1"
         yield "cumulative2"
         yield "[DONE]"
 
-    mock_engine.generate.return_value = stream_gen()
+    mock_engine.generate_async.return_value = async_gen()
     monkeypatch.setattr("astrai.inference.server._engine", mock_engine)
     response = client.post(
         "/v1/chat/completions",