Like SpiceDB bulk delete, we need a mass cleanup mechanism for Inventory. We cannot rely on manual psql access or the Kessel Debug Pod for bulk deletion in Prod during an emergency. 

One potential solution is utilizing a kube job and custom cobra command built on top of inventory's existing gorm usage

• The job must perform deletes in batches (e.g., 5,000 records at a time) with short sleeps in between to allow the database to breathe and autovacuum to catch up.
• We can’t use OFFSET/LIMIT for deletion loops (it gets slower as it gets to later records). We should use keyset pagination.
• Package this logic into a cobra command, and then a Kubernetes Job that we/AppSRE can trigger.