fix: resolve order queue loss issue with retry mechanism and idempotency

- Fix task error handling: return actual errors instead of nil to enable retry
- Add idempotency check: skip processing for already finished orders
- Extend temp order cache: increase from 15 minutes to 24 hours
- Configure retry policy: add MaxRetry(5) for all payment callbacks (Epay, Alipay, Stripe)

This fixes the critical issue where paid orders were being lost due to:
1. Failed tasks being marked as successful and deleted from queue
2. Temporary order info expiring before queue processing
3. No retry mechanism for transient failures

Changes:
- queue/logic/order/activateOrderLogic.go: Fix error returns and add idempotency
- internal/logic/public/portal/purchaseLogic.go: Extend cache to 24 hours
- internal/logic/notify/*NotifyLogic.go: Add retry configuration

internal/logic/notify/alipayNotifyLogic.go

@@ -82,7 +82,7 @@ func (l *AlipayNotifyLogic) AlipayNotify(r *http.Request) error {
 			l.Logger.Error("[AlipayNotify] Marshal payload failed", logger.Field("error", err.Error()))
 			return err
 		}
-		task := asynq.NewTask(types.ForthwithActivateOrder, bytes)
+		task := asynq.NewTask(types.ForthwithActivateOrder, bytes, asynq.MaxRetry(5))
 		taskInfo, err := l.svcCtx.Queue.EnqueueContext(l.ctx, task)
 		if err != nil {
 			l.Logger.Error("[AlipayNotify] Enqueue task failed", logger.Field("error", err.Error()))