perf: 优化请求鉴权链路并批量化统计/调度查询

- 为 Pipeline/Context 增加按需读取请求体能力，支持 async 懒加载 JSON body
- 为 chat/cli/video/claude/openai-cli 适配器关闭默认预读，减少无效 body 读取与超时风险
- 将本地登录、JWT 用户加载、API Key 鉴权迁移到线程池隔离会话执行，避免阻塞事件循环
- 为 API Key 鉴权返回结构化余额结果，并在主请求会话中重新绑定 user/api_key 后再校验状态、过期和锁定信息
- 为 management/user token 前缀认证引入独立会话与结果回绑，避免跨会话对象写入失效

- 为 Usage 余额检查补充结构化返回，统一透出 remaining 与欠费/不可用文案映射
- 为用户与管理端活跃请求查询增加 maintain_status 开关，避免轮询指定 id 时误触发状态修复
- 重写 user_me usage 汇总逻辑，支持 group_by=None 的粗粒度聚合
- 修正 provider 维度成功率与平均响应时间统计，基于 success_count 和成功响应耗时汇总计算
- 前端 Usage 轮询由 setInterval 改为串行 setTimeout，避免并发轮询叠加

- 为 StatsAggregator 增加按本地日期批量计算百分位能力，替代逐天 fan-out 查询
- 为混合统计查询合并连续实时日期区间，并批量读取 StatsDaily，减少逐日查询次数
- 为用户日统计增加批量聚合入口，替代逐用户循环聚合
- 为系统配置导出改用 selectinload 预加载 provider 关联数据，减少 N+1 查询
- 为管理员用户列表增加钱包批量查询，避免逐用户回表

- 为调度器增加 provider 轻量引用预过滤，先按 allowed_providers 缩小范围再加载完整 provider 图
- 为 CandidateBuilder 增加 provider refs/provider_ids 查询能力，保留分页顺序
- 为模型缓存增加 provider_model_mappings 索引缓存与 model_mappings 规则缓存，减少重复全量扫描
- 为请求候选中间态改为 flush/batch commit，降低 pending/streaming 状态切换的事务往返
- 为钱包访问结果补充 balance_snapshot，并抽取余额快照复用逻辑

- 补充 pipeline、auth、admin users、user_me usage、stats aggregator、model cache、
  scheduler、wallet、request candidate 等回归与契约测试

tests/services/test_stats_aggregator_optimization.py

@@ -0,0 +1,182 @@
+from __future__ import annotations
+
+from datetime import date, datetime, time, timedelta, timezone
+from types import SimpleNamespace
+from typing import Any, cast
+
+import pytest
+
+from src.models.database import StatsDaily, StatsUserDaily
+from src.services.system.stats_aggregator import (
+    AggregatedStats,
+    StatsAggregatorService,
+    query_stats_hybrid,
+)
+from src.services.system.time_range import TimeRangeParams
+
+
+class _FakeQuery:
+    def __init__(self, *, all_result: list[Any] | None = None) -> None:
+        self._all_result = all_result if all_result is not None else []
+
+    def filter(self, *_args: object, **_kwargs: object) -> _FakeQuery:
+        return self
+
+    def group_by(self, *_args: object, **_kwargs: object) -> _FakeQuery:
+        return self
+
+    def all(self) -> list[Any]:
+        return self._all_result
+
+
+class _HybridQuerySession:
+    def __init__(self, stats_daily_rows: list[SimpleNamespace]) -> None:
+        self._stats_daily_rows = stats_daily_rows
+        self.stats_daily_query_count = 0
+
+    def query(self, entity: object) -> _FakeQuery:
+        if entity is StatsDaily:
+            self.stats_daily_query_count += 1
+            return _FakeQuery(all_result=self._stats_daily_rows)
+        raise AssertionError(f"Unexpected query entity: {entity}")
+
+
+class _BatchUserStatsSession:
+    def __init__(
+        self, existing_rows: list[StatsUserDaily], aggregated_rows: list[SimpleNamespace]
+    ) -> None:
+        self._responses: list[list[Any]] = [list(existing_rows), list(aggregated_rows)]
+        self.added: list[StatsUserDaily] = []
+        self.commit_count = 0
+
+    def query(self, *_entities: object) -> _FakeQuery:
+        if not self._responses:
+            raise AssertionError("Unexpected extra query")
+        return _FakeQuery(all_result=self._responses.pop(0))
+
+    def add(self, row: StatsUserDaily) -> None:
+        self.added.append(row)
+
+    def commit(self) -> None:
+        self.commit_count += 1
+
+
+def test_query_stats_hybrid_batches_statsdaily_lookup_and_merges_realtime_ranges(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    today = datetime.now(timezone.utc).date()
+    historical_cached_day = today - timedelta(days=4)
+    historical_missing_day = today - timedelta(days=3)
+    realtime_day = today
+
+    cached_row = SimpleNamespace(
+        date=datetime.combine(historical_cached_day, time.min, tzinfo=timezone.utc),
+        total_requests=10,
+        success_requests=9,
+        error_requests=1,
+        input_tokens=100,
+        output_tokens=50,
+        cache_creation_tokens=5,
+        cache_read_tokens=3,
+        cache_creation_cost=1.2,
+        cache_read_cost=0.8,
+        total_cost=3.5,
+        actual_total_cost=3.0,
+        avg_response_time_ms=200.0,
+    )
+    db = _HybridQuerySession(stats_daily_rows=[cached_row])
+
+    calls: list[tuple[datetime, datetime]] = []
+
+    def _fake_aggregate_usage_range(
+        _db: object,
+        start_utc: datetime,
+        end_utc: datetime,
+        filters: object | None = None,  # noqa: ARG001
+    ) -> AggregatedStats:
+        calls.append((start_utc, end_utc))
+        return AggregatedStats(total_requests=1, success_requests=1)
+
+    class _FakeParams:
+        def get_complete_utc_dates(self) -> tuple[list[date], None, None]:
+            return [historical_cached_day, historical_missing_day, realtime_day], None, None
+
+    monkeypatch.setattr(
+        "src.services.system.stats_aggregator.aggregate_usage_range",
+        _fake_aggregate_usage_range,
+    )
+
+    result = query_stats_hybrid(cast(Any, db), cast(Any, _FakeParams()))
+
+    assert db.stats_daily_query_count == 1
+    assert calls == [
+        (
+            datetime.combine(historical_missing_day, time.min, tzinfo=timezone.utc),
+            datetime.combine(
+                historical_missing_day + timedelta(days=1), time.min, tzinfo=timezone.utc
+            ),
+        ),
+        (
+            datetime.combine(realtime_day, time.min, tzinfo=timezone.utc),
+            datetime.combine(realtime_day + timedelta(days=1), time.min, tzinfo=timezone.utc),
+        ),
+    ]
+    assert result.total_requests == 12
+    assert result.success_requests == 11
+
+
+def test_aggregate_user_daily_stats_batch_updates_all_users_in_two_queries() -> None:
+    target_day = datetime(2026, 3, 1, tzinfo=timezone.utc)
+    aggregated_rows = [
+        SimpleNamespace(
+            user_id="user-1",
+            username="alice",
+            total_requests=4,
+            error_requests=1,
+            input_tokens=20,
+            output_tokens=8,
+            cache_creation_tokens=2,
+            cache_read_tokens=1,
+            total_cost=1.5,
+        )
+    ]
+    db = _BatchUserStatsSession(existing_rows=[], aggregated_rows=aggregated_rows)
+
+    result = StatsAggregatorService.aggregate_user_daily_stats_batch(
+        cast(Any, db),
+        target_day,
+        ["user-1", "user-2"],
+        commit=True,
+    )
+
+    assert len(result) == 2
+    assert db.commit_count == 1
+    assert len(db.added) == 2
+
+    user_one = next(row for row in result if row.user_id == "user-1")
+    user_two = next(row for row in result if row.user_id == "user-2")
+
+    assert user_one.username == "alice"
+    assert user_one.total_requests == 4
+    assert user_one.success_requests == 3
+    assert user_one.total_cost == 1.5
+
+    assert user_two.total_requests == 0
+    assert user_two.success_requests == 0
+    assert user_two.error_requests == 0
+    assert user_two.total_cost == 0.0
+
+
+def test_compute_percentiles_by_local_day_returns_sqlite_fallback_without_queries() -> None:
+    db = SimpleNamespace(bind=SimpleNamespace(dialect=SimpleNamespace(name="sqlite")))
+    time_range = TimeRangeParams(
+        start_date=date(2026, 3, 1),
+        end_date=date(2026, 3, 3),
+        timezone="Asia/Singapore",
+    )
+
+    result = StatsAggregatorService.compute_percentiles_by_local_day(cast(Any, db), time_range)
+
+    assert [row["date"] for row in result] == ["2026-03-01", "2026-03-02", "2026-03-03"]
+    assert all(row["p50_response_time_ms"] is None for row in result)
+    assert all(row["p50_first_byte_time_ms"] is None for row in result)