feat: squash merge all changes from develop-0.1.75

Squash of 124 commits from the legacy develop branch (develop-0.1.75)
onto a clean v0.1.75 upstream base, to simplify future upstream merges.

Key changes included:
- Refactor scope-level rate limiting to model-level rate limiting
- Antigravity gateway service improvements (smart retry, error policy)
- Digest session store (flat cache replacing Trie-based store)
- Client disconnect detection during streaming
- Gemini messages compatibility service enhancements
- Scheduler shuffle for thundering herd prevention
- Session hash generation improvements
- Frontend customizations (WeChat service, HomeView, etc.)
- Ops monitoring scope cleanup

backend/internal/service/openai_gateway_service.go

@@ -580,10 +580,6 @@ func (s *OpenAIGatewayService) SelectAccountWithLoadAwareness(ctx context.Contex
 			}
 		}
 	} else {
-		type accountWithLoad struct {
-			account  *Account
-			loadInfo *AccountLoadInfo
-		}
 		var available []accountWithLoad
 		for _, acc := range candidates {
 			loadInfo := loadMap[acc.ID]
@@ -618,6 +614,7 @@ func (s *OpenAIGatewayService) SelectAccountWithLoadAwareness(ctx context.Contex
 					return a.account.LastUsedAt.Before(*b.account.LastUsedAt)
 				}
 			})
+			shuffleWithinSortGroups(available)
 
 			for _, item := range available {
 				result, err := s.tryAcquireAccountSlot(ctx, item.account.ID, item.account.Concurrency)