Cache keys in database

[nbgl]

Currently we run key derivation every time we begin hashing a batch. This is expensive. Key derivation should only happen once.

But maybe it shouldn’t happen in post_pii, so we don’t make the client wait a long time for a response. This is something we can benchmark. It should be possible to respond to the client and perform the key derivation in Celery before doing hashing.

Aha! Link: https://csiro.aha.io/features/ANONLINK-91