refactor(ws): 用 ResponsesTurnState 收敛连接 turn 状态

评审第 2 条：BoundResponsesConnection 用 response_in_flight、active_turn、
active_response_create 三个可独立变化的字段编码同一件事，8 种组合里只有 3 种
合法，非法组合只能靠调用点的 if 和「记得同时改另外两个字段」来避免。

三字段合并为一个 ResponsesTurnState：

  Idle                                  没有进行中的 logical turn
  Responding { logical, attempt }        logical 与 attempt 必须同时存在
  Replanning { logical }                 attempt 已取走去结算/重绑，logical 仍在

Replanning 不是新概念：配额透明重试期间现状就处于这个状态，只是靠
Option::take 意外得到。转换只能走 begin / detach_attempt / resume / end，
response_in_flight 与「是否接受新 response.create」都由变体推导。

由此消除的运行时不变量（原来全靠调用点自觉）：
- 有 attempt 必有 logical turn
- response_in_flight 与 attempt 同生共死（原来 client 写失败后
  active_turn=None 而 response_in_flight 仍为 true）
- logical turn 结束时必须清 attempt：原来 `active_response_create = None`
  在 connection.rs 里手写 13 处，漏一处就残留；现在只有 end() 一个出口
- 上游绑定返回的连接不再自带 response_in_flight=true 的半成品状态

同时删除 update_response_in_flight：Started 帧把已经是 true 的字段再设一次，
Close 帧因为没有解析出的 frame 而根本不触发，是纯冗余写；它在 Idle 态收到
Started 帧时还会把 response_in_flight 置真，从而永久阻塞后续 response.create。

行为等价。ActiveResponsesWebSocketRequest 改名 LogicalTurn 并随状态机移入
新的 turn_state.rs；状态机对 attempt 类型泛型化，测试用轻量替身驱动同一套
转换逻辑，无需 AppState 或真实 socket。

apps/aether-gateway/src/handlers/proxy/websocket/responses/state.rs

@@ -4,14 +4,12 @@
 //! connection may survive many `response.create` turns, while the turn
 //! lifecycle and upstream binding are replaced independently.
 
-use serde_json::Value;
 use std::collections::{BTreeMap, BTreeSet};
 use tokio::task::JoinHandle;
 
 use super::adapter::{ResponsesWebSocketDrainDirective, ResponsesWebSocketProtocolAdapter};
 use super::binding::UpstreamBindingIdentity;
-use super::lifecycle::ActiveResponsesWebSocketTurn;
-use super::request::response_create_has_previous_response_id;
+use super::turn_state::ResponsesTurnState;
 use crate::ai_serving::{AiExecutionDecision, ResponsesWebSocketBodyNormalization};
 
 const EXHAUSTED_KEY_EXCLUSION_FALLBACK_SECONDS: u64 = 300;
@@ -22,15 +20,14 @@ pub(super) struct BoundResponsesConnection {
     pub(super) adapter: &'static dyn ResponsesWebSocketProtocolAdapter,
     pub(super) client_model: String,
     pub(super) provider_model: String,
-    pub(super) response_in_flight: bool,
     pub(super) decision_template: AiExecutionDecision,
     /// Reproduces this binding's provider-body normalization for continuation
     /// turns, which must not re-enter the planner. Replaced whenever the
     /// binding or its decision is replaced.
     pub(super) body_normalization: ResponsesWebSocketBodyNormalization,
     pub(super) binding_identity: UpstreamBindingIdentity,
-    pub(super) active_turn: Option<ActiveResponsesWebSocketTurn>,
-    pub(super) active_response_create: Option<ActiveResponsesWebSocketRequest>,
+    /// 这条连接上「有没有正在进行的 logical turn」的唯一事实来源。
+    pub(super) turn_state: ResponsesTurnState,
     pub(super) next_turn_index: u64,
     pub(super) upstream_response_headers: BTreeMap<String, String>,
     pub(super) pending_adapter_drain: Option<ResponsesWebSocketDrainDirective>,
@@ -101,41 +98,3 @@ impl ExhaustedResponsesWebSocketExclusions {
     }
 }
 
-#[derive(Debug, Clone)]
-pub(super) struct ActiveResponsesWebSocketRequest {
-    pub(super) client_event: Value,
-    pub(super) turn_index: u64,
-    pub(super) logical_turn_id: String,
-    pub(super) turn_attempt: u32,
-    pub(super) retry_attempted: bool,
-    pub(super) retry_unsafe_reason: Option<&'static str>,
-}
-
-impl ActiveResponsesWebSocketRequest {
-    pub(super) fn new(client_event: Value, turn_index: u64, logical_turn_id: String) -> Self {
-        Self {
-            client_event,
-            turn_index,
-            logical_turn_id,
-            turn_attempt: 1,
-            retry_attempted: false,
-            retry_unsafe_reason: None,
-        }
-    }
-
-    pub(super) fn quota_retry_block_reason(&self) -> Option<&'static str> {
-        if self.retry_attempted {
-            Some("quota_retry_already_attempted")
-        } else if let Some(reason) = self.retry_unsafe_reason {
-            Some(reason)
-        } else if response_create_has_previous_response_id(&self.client_event) {
-            Some("previous_response_id")
-        } else {
-            None
-        }
-    }
-
-    pub(super) fn mark_retry_unsafe(&mut self, reason: &'static str) {
-        self.retry_unsafe_reason.get_or_insert(reason);
-    }
-}