Implement caching?

We currently have a sort of in-memory cache by virtue of how Tracks work internally, but that's not at all set up for the type of random access that would be required by FETCH. Consider adding a minimal in-memory or on-disk cache component so we can serve FETCH requests locally and also locally fulfill Joining Fetch requests.

Start by tracing the existing FETCH and Joining Fetch handling in the repository, then compare it with how Tracks currently provide in-memory access. Define the cache component's scope and storage behavior before implementation. Done means local FETCH requests and Joining Fetch requests can be fulfilled through the cache, with appropriate coverage for both paths.

Written by the indexing model from the issue text.