fix(网关): 修复账号选择中的调度器快照延迟问题

## 问题描述
调度器快照更新存在0.5-1秒的延迟（Outbox轮询间隔），导致在账号被限流或过载后的短时间窗口内，
可能仍会被选中，造成请求失败。

## 根本原因
账号选择逻辑依赖调度器快照（listSchedulableAccounts），但快照更新有延迟：
- Outbox轮询: 每1秒检查一次变更事件
- 全量重建: 每300秒重建一次
- 时间窗口: 账号状态变更后0.5-1秒内，快照可能未更新

## 解决方案
在账号选择循环中添加IsSchedulable()实时检查，作为第二道防线：
1. 第一道防线: 调度器快照过滤（可能有延迟）
2. 第二道防线: IsSchedulable()实时检查（本次修复）

IsSchedulable()会检查：
- RateLimitResetAt: 限流重置时间
- OverloadUntil: 过载持续时间
- TempUnschedulableUntil: 临时不可调度时间
- Status: 账号状态
- Schedulable: 可调度标志

## 修改范围
### OpenAI Gateway Service
- SelectAccountForModelWithExclusions: 添加IsSchedulable()检查
- SelectAccountWithLoadAwareness: 添加IsSchedulable()检查

### Gateway Service (Claude/Gemini/Antigravity)
- 负载感知选择候选账号筛选: 添加IsSchedulable()检查
- selectAccountForModelWithPlatform: 添加IsSchedulable()检查
- selectAccountWithMixedScheduling: 添加IsSchedulable()检查

### 测试用例
- OpenAI: 添加2个测试用例验证限流账号过滤
- Gateway: 添加2个测试用例验证限流和过载账号过滤

### 其他修复
- ops_repo_preagg.go: 修复platform为NULL时的聚合问题

## 测试结果
所有单元测试通过 ✅

backend/internal/service/openai_gateway_service_test.go

@@ -3,6 +3,7 @@ package service
 import (
 	"bufio"
 	"bytes"
+	"context"
 	"errors"
 	"io"
 	"net/http"
@@ -15,6 +16,129 @@ import (
 	"github.com/gin-gonic/gin"
 )
 
+type stubOpenAIAccountRepo struct {
+	AccountRepository
+	accounts []Account
+}
+
+func (r stubOpenAIAccountRepo) ListSchedulableByGroupIDAndPlatform(ctx context.Context, groupID int64, platform string) ([]Account, error) {
+	return append([]Account(nil), r.accounts...), nil
+}
+
+func (r stubOpenAIAccountRepo) ListSchedulableByPlatform(ctx context.Context, platform string) ([]Account, error) {
+	return append([]Account(nil), r.accounts...), nil
+}
+
+type stubConcurrencyCache struct {
+	ConcurrencyCache
+}
+
+func (c stubConcurrencyCache) AcquireAccountSlot(ctx context.Context, accountID int64, maxConcurrency int, requestID string) (bool, error) {
+	return true, nil
+}
+
+func (c stubConcurrencyCache) ReleaseAccountSlot(ctx context.Context, accountID int64, requestID string) error {
+	return nil
+}
+
+func (c stubConcurrencyCache) GetAccountsLoadBatch(ctx context.Context, accounts []AccountWithConcurrency) (map[int64]*AccountLoadInfo, error) {
+	out := make(map[int64]*AccountLoadInfo, len(accounts))
+	for _, acc := range accounts {
+		out[acc.ID] = &AccountLoadInfo{AccountID: acc.ID, LoadRate: 0}
+	}
+	return out, nil
+}
+
+func TestOpenAISelectAccountWithLoadAwareness_FiltersUnschedulable(t *testing.T) {
+	now := time.Now()
+	resetAt := now.Add(10 * time.Minute)
+	groupID := int64(1)
+
+	rateLimited := Account{
+		ID:               1,
+		Platform:         PlatformOpenAI,
+		Type:             AccountTypeAPIKey,
+		Status:           StatusActive,
+		Schedulable:      true,
+		Concurrency:      1,
+		Priority:         0,
+		RateLimitResetAt: &resetAt,
+	}
+	available := Account{
+		ID:          2,
+		Platform:    PlatformOpenAI,
+		Type:        AccountTypeAPIKey,
+		Status:      StatusActive,
+		Schedulable: true,
+		Concurrency: 1,
+		Priority:    1,
+	}
+
+	svc := &OpenAIGatewayService{
+		accountRepo:        stubOpenAIAccountRepo{accounts: []Account{rateLimited, available}},
+		concurrencyService: NewConcurrencyService(stubConcurrencyCache{}),
+	}
+
+	selection, err := svc.SelectAccountWithLoadAwareness(context.Background(), &groupID, "", "gpt-5.2", nil)
+	if err != nil {
+		t.Fatalf("SelectAccountWithLoadAwareness error: %v", err)
+	}
+	if selection == nil || selection.Account == nil {
+		t.Fatalf("expected selection with account")
+	}
+	if selection.Account.ID != available.ID {
+		t.Fatalf("expected account %d, got %d", available.ID, selection.Account.ID)
+	}
+	if selection.ReleaseFunc != nil {
+		selection.ReleaseFunc()
+	}
+}
+
+func TestOpenAISelectAccountWithLoadAwareness_FiltersUnschedulableWhenNoConcurrencyService(t *testing.T) {
+	now := time.Now()
+	resetAt := now.Add(10 * time.Minute)
+	groupID := int64(1)
+
+	rateLimited := Account{
+		ID:               1,
+		Platform:         PlatformOpenAI,
+		Type:             AccountTypeAPIKey,
+		Status:           StatusActive,
+		Schedulable:      true,
+		Concurrency:      1,
+		Priority:         0,
+		RateLimitResetAt: &resetAt,
+	}
+	available := Account{
+		ID:          2,
+		Platform:    PlatformOpenAI,
+		Type:        AccountTypeAPIKey,
+		Status:      StatusActive,
+		Schedulable: true,
+		Concurrency: 1,
+		Priority:    1,
+	}
+
+	svc := &OpenAIGatewayService{
+		accountRepo: stubOpenAIAccountRepo{accounts: []Account{rateLimited, available}},
+		// concurrencyService is nil, forcing the non-load-batch selection path.
+	}
+
+	selection, err := svc.SelectAccountWithLoadAwareness(context.Background(), &groupID, "", "gpt-5.2", nil)
+	if err != nil {
+		t.Fatalf("SelectAccountWithLoadAwareness error: %v", err)
+	}
+	if selection == nil || selection.Account == nil {
+		t.Fatalf("expected selection with account")
+	}
+	if selection.Account.ID != available.ID {
+		t.Fatalf("expected account %d, got %d", available.ID, selection.Account.ID)
+	}
+	if selection.ReleaseFunc != nil {
+		selection.ReleaseFunc()
+	}
+}
+
 func TestOpenAIStreamingTimeout(t *testing.T) {
 	gin.SetMode(gin.TestMode)
 	cfg := &config.Config{