Improve WAL-E alert for 'WALEBackupDelayed'

During our recent failover production#637 (closed), the alert WALEBackupDelayed fired. This was technically a false alarm. Wal-E backups continued to work, the data that prometheus was looking at was correct, but fired when another server had begun processing uploads to S3.

https://prometheus.gprd.gitlab.net/graph?g0.range_input=2d&g0.expr=gitlab_com%3Alast_wale_backup_age_in_seconds%20&g0.tab=0

Improve this alert to reduce alert fatigue during incidents. This alert introduced noise into an already stressful situation.