Cache DB should be aware of the concept of substores

[andaaron]

Currently the cache DB is separate per store (default store and substores each have their own separate cache DB files).
Ideally we should not have to manage this many DBs.

The reason they are separate is that we don't want to dedupe files placed on separate stores (for example default store and a substore). This is because they stores may be on separate partitions, which makes hardlinks unusable in this case. Or a store may be local and another S3.

If we had a concept of substore in the cache DB logic, and it would be capable of not deduping objects on separate stores, we could reuse the same DB.

If we used the same DB for all cache, then we could potentially use the same DB file also for repo DB and user DB, and manage a single DB.