Consider removing hard deletion of clusters and their data

[angstwad]

Consider putting it into a disabled state with a model flag

Querying of cluster model state via API uses disabled=true query arg