feat(antigravity): comprehensive enhancements - model mapping, rate limiting, scheduling & ops

Key changes:
- Upgrade model mapping: Opus 4.5 → Opus 4.6-thinking with precise matching
- Unified rate limiting: scope-level → model-level with Redis snapshot sync
- Load-balanced scheduling by call count with smart retry mechanism
- Force cache billing support
- Model identity injection in prompts with leak prevention
- Thinking mode auto-handling (max_tokens/budget_tokens fix)
- Frontend: whitelist mode toggle, model mapping validation, status indicators
- Gemini session fallback with Redis Trie O(L) matching
- Ops: enhanced concurrency monitoring, account availability, retry logic
- Migration scripts: 049-051 for model mapping unification

backend/internal/service/gemini_messages_compat_service.go

@@ -200,7 +200,7 @@ func (s *GeminiMessagesCompatService) tryStickySessionHit(
 
 	// 检查账号是否需要清理粘性会话
 	// Check if sticky session should be cleared
-	if shouldClearStickySession(account) {
+	if shouldClearStickySession(account, requestedModel) {
 		_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), cacheKey)
 		return nil
 	}
@@ -230,7 +230,7 @@ func (s *GeminiMessagesCompatService) isAccountUsableForRequest(
 ) bool {
 	// 检查模型调度能力
 	// Check model scheduling capability
-	if !account.IsSchedulableForModel(requestedModel) {
+	if !account.IsSchedulableForModelWithContext(ctx, requestedModel) {
 		return false
 	}
 
@@ -1498,28 +1498,6 @@ func (s *GeminiMessagesCompatService) writeGeminiMappedError(c *gin.Context, acc
 		log.Printf("[Gemini] upstream error %d: %s", upstreamStatus, truncateForLog(body, s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes))
 	}
 
-	if status, errType, errMsg, matched := applyErrorPassthroughRule(
-		c,
-		PlatformGemini,
-		upstreamStatus,
-		body,
-		http.StatusBadGateway,
-		"upstream_error",
-		"Upstream request failed",
-	); matched {
-		c.JSON(status, gin.H{
-			"type":  "error",
-			"error": gin.H{"type": errType, "message": errMsg},
-		})
-		if upstreamMsg == "" {
-			upstreamMsg = errMsg
-		}
-		if upstreamMsg == "" {
-			return fmt.Errorf("upstream error: %d (passthrough rule matched)", upstreamStatus)
-		}
-		return fmt.Errorf("upstream error: %d (passthrough rule matched) message=%s", upstreamStatus, upstreamMsg)
-	}
-
 	var statusCode int
 	var errType, errMsg string
 
@@ -2658,7 +2636,9 @@ func ParseGeminiRateLimitResetTime(body []byte) *int64 {
 					if meta, ok := dm["metadata"].(map[string]any); ok {
 						if v, ok := meta["quotaResetDelay"].(string); ok {
 							if dur, err := time.ParseDuration(v); err == nil {
-								ts := time.Now().Unix() + int64(dur.Seconds())
+								// Use ceil to avoid undercounting fractional seconds (e.g. 10.1s should not become 10s),
+								// which can affect scheduling decisions around thresholds (like 10s).
+								ts := time.Now().Unix() + int64(math.Ceil(dur.Seconds()))
 								return &ts
 							}
 						}