Improve caching strategy

[jacobwegner]

(Builds on #1)

In the future, the project should consider switching to the RedisCache backend and configuring Redis with the allkeys-lru eviction policy.

This would allow a constrained cache footprint (rather than the open-ended cache size configured in #1).

I don't expect cache growth to be a problem with the current configuration, as the Kubernetes deployment for https://scaife.perseus.org has plenty of disk space available.