fix(pool): 配额倒计时归零后进度条与列表恢复显示 100%

- 后端读取层 provider_key_status_snapshot_payload 复用调度侧同一判定
  provider_pool_reset_deadline_elapsed，对已到期的 Codex 配额窗口归一化为
  used_ratio=0 / remaining_ratio=1，并清除窗口级耗尽标记；跳过
  window_minutes=0 与无用量观测的窗口
- 前端展示层在倒计时归零（isExpired）时兜底显示 100%，并隐藏重置前的旧用量
  文本，保证归零瞬间即时恢复，与后端读取口径、调度口径一致
- 补充后端 2 个单测（到期归一化、遗留耗尽状态清理）与前端 1 个回归用例
  （负向验证可复现旧行为）

验证：aether-provider-pool 71/71；aether-gateway lib 回归通过（仅 3 个依赖
PostgreSQL 服务端的用例因本机环境缺失失败，与本改动无关）；前端 222 文件 /
1718 用例全部通过；clippy -D warnings、cargo fmt、vue-tsc、eslint 均通过。

crates/aether-provider/pool/src/lib.rs

@@ -43,7 +43,7 @@ pub use quota::{
     provider_pool_key_model_quota_hard_blocked, provider_pool_key_quota_hard_blocked,
     provider_pool_key_scheduling_label, provider_pool_member_quota_snapshot,
     provider_pool_quota_metadata_provider_type, provider_pool_quota_metadata_updated_at,
-    provider_pool_quota_snapshot_updated_at,
+    provider_pool_quota_snapshot_updated_at, provider_pool_reset_deadline_elapsed,
 };
 pub use quota_refresh::ProviderPoolQuotaRequestSpec;
 pub use service::ProviderPoolService;

crates/aether-provider/pool/src/quota.rs

@@ -857,7 +857,14 @@ fn provider_pool_reset_deadline_unix_secs(
         })
 }
 
-pub(crate) fn provider_pool_reset_deadline_elapsed(
+/// Whether the quota window's reset deadline has already passed.
+///
+/// The deadline is taken from `reset_at`/`next_reset_at` when present, otherwise
+/// derived from `reset_seconds`/`reset_after_seconds` anchored at the window (or
+/// fallback) observation time. Scheduling ignores exhausted windows once this is
+/// true; read paths can reuse the same predicate so the displayed quota matches
+/// the scheduling decision after a reset.
+pub fn provider_pool_reset_deadline_elapsed(
     item: &Map<String, Value>,
     fallback_observed_at: Option<u64>,
     now_unix_secs: u64,