feat(gateway): 实现负载感知的账号调度优化

- 新增调度配置：粘性会话排队、兜底排队、负载计算、槽位清理
- 实现账号级等待队列和批量负载查询（Redis Lua 脚本）
- 三层选择策略：粘性会话优先 → 负载感知选择 → 兜底排队
- 后台定期清理过期槽位，防止资源泄漏
- 集成到所有网关处理器（Claude/Gemini/OpenAI）

backend/internal/config/config_test.go

@@ -1,6 +1,11 @@
 package config
 
-import "testing"
+import (
+	"testing"
+	"time"
+
+	"github.com/spf13/viper"
+)
 
 func TestNormalizeRunMode(t *testing.T) {
 	tests := []struct {
@@ -21,3 +26,45 @@ func TestNormalizeRunMode(t *testing.T) {
 		}
 	}
 }
+
+func TestLoadDefaultSchedulingConfig(t *testing.T) {
+	viper.Reset()
+
+	cfg, err := Load()
+	if err != nil {
+		t.Fatalf("Load() error: %v", err)
+	}
+
+	if cfg.Gateway.Scheduling.StickySessionMaxWaiting != 3 {
+		t.Fatalf("StickySessionMaxWaiting = %d, want 3", cfg.Gateway.Scheduling.StickySessionMaxWaiting)
+	}
+	if cfg.Gateway.Scheduling.StickySessionWaitTimeout != 45*time.Second {
+		t.Fatalf("StickySessionWaitTimeout = %v, want 45s", cfg.Gateway.Scheduling.StickySessionWaitTimeout)
+	}
+	if cfg.Gateway.Scheduling.FallbackWaitTimeout != 30*time.Second {
+		t.Fatalf("FallbackWaitTimeout = %v, want 30s", cfg.Gateway.Scheduling.FallbackWaitTimeout)
+	}
+	if cfg.Gateway.Scheduling.FallbackMaxWaiting != 100 {
+		t.Fatalf("FallbackMaxWaiting = %d, want 100", cfg.Gateway.Scheduling.FallbackMaxWaiting)
+	}
+	if !cfg.Gateway.Scheduling.LoadBatchEnabled {
+		t.Fatalf("LoadBatchEnabled = false, want true")
+	}
+	if cfg.Gateway.Scheduling.SlotCleanupInterval != 30*time.Second {
+		t.Fatalf("SlotCleanupInterval = %v, want 30s", cfg.Gateway.Scheduling.SlotCleanupInterval)
+	}
+}
+
+func TestLoadSchedulingConfigFromEnv(t *testing.T) {
+	viper.Reset()
+	t.Setenv("GATEWAY_SCHEDULING_STICKY_SESSION_MAX_WAITING", "5")
+
+	cfg, err := Load()
+	if err != nil {
+		t.Fatalf("Load() error: %v", err)
+	}
+
+	if cfg.Gateway.Scheduling.StickySessionMaxWaiting != 5 {
+		t.Fatalf("StickySessionMaxWaiting = %d, want 5", cfg.Gateway.Scheduling.StickySessionMaxWaiting)
+	}
+}