perf: 优化负载感知调度的准确性和响应速度

基于 Codex 审查建议的性能优化。

负载批量查询优化：
- getAccountsLoadBatchScript 添加过期槽位清理
- 使用 ZREMRANGEBYSCORE 在计数前清理过期条目
- 防止过期槽位导致负载率计算偏高
- 提升负载感知调度的准确性

等待循环优化：
- waitForSlotWithPingTimeout 添加立即获取尝试
- 避免不必要的 initialBackoff 延迟
- 低负载场景下减少响应延迟

测试改进：
- 取消跳过 TestGetAccountsLoadBatch 集成测试
- 过期槽位清理应该修复了 CI 中的计数问题

影响：
- 更准确的负载感知调度决策
- 更快的槽位获取响应
- 更好的测试覆盖率

backend/internal/handler/gateway_helper.go

@@ -144,6 +144,21 @@ func (h *ConcurrencyHelper) waitForSlotWithPingTimeout(c *gin.Context, slotType
 	ctx, cancel := context.WithTimeout(c.Request.Context(), timeout)
 	defer cancel()
 
+	// Try immediate acquire first (avoid unnecessary wait)
+	var result *service.AcquireResult
+	var err error
+	if slotType == "user" {
+		result, err = h.concurrencyService.AcquireUserSlot(ctx, id, maxConcurrency)
+	} else {
+		result, err = h.concurrencyService.AcquireAccountSlot(ctx, id, maxConcurrency)
+	}
+	if err != nil {
+		return nil, err
+	}
+	if result.Acquired {
+		return result.ReleaseFunc, nil
+	}
+
 	// Determine if ping is needed (streaming + ping format defined)
 	needPing := isStream && h.pingFormat != ""
 

backend/internal/repository/concurrency_cache.go

@@ -151,11 +151,17 @@ var (
 			return 1
 		`)
 
-	// getAccountsLoadBatchScript - batch load query (read-only)
-	// ARGV[1] = slot TTL (seconds, retained for compatibility)
+	// getAccountsLoadBatchScript - batch load query with expired slot cleanup
+	// ARGV[1] = slot TTL (seconds)
 	// ARGV[2..n] = accountID1, maxConcurrency1, accountID2, maxConcurrency2, ...
 	getAccountsLoadBatchScript = redis.NewScript(`
 			local result = {}
+			local slotTTL = tonumber(ARGV[1])
+
+			-- Get current server time
+			local timeResult = redis.call('TIME')
+			local nowSeconds = tonumber(timeResult[1])
+			local cutoffTime = nowSeconds - slotTTL
 
 			local i = 2
 			while i <= #ARGV do
@@ -163,6 +169,9 @@ var (
 				local maxConcurrency = tonumber(ARGV[i + 1])
 
 				local slotKey = 'concurrency:account:' .. accountID
+
+				-- Clean up expired slots before counting
+				redis.call('ZREMRANGEBYSCORE', slotKey, '-inf', cutoffTime)
 				local currentConcurrency = redis.call('ZCARD', slotKey)
 
 				local waitKey = 'wait:account:' .. accountID

backend/internal/repository/concurrency_cache_integration_test.go

@@ -275,7 +275,6 @@ func (s *ConcurrencyCacheSuite) TestGetUserConcurrency_Missing() {
 }
 
 func (s *ConcurrencyCacheSuite) TestGetAccountsLoadBatch() {
-	s.T().Skip("TODO: Fix this test - CurrentConcurrency returns 0 instead of expected value in CI")
 	// Setup: Create accounts with different load states
 	account1 := int64(100)
 	account2 := int64(101)