Add fallback in case remote codelists are inaccessible

[jpmckinney]

5427afc fixed several bugs, by caching codelists in-memory for a day.

However, if a remote URL is inaccessible, then an error is raised.

We could implement a fallback, where the codelists are also written to disk. Whenever the in-memory cache expires, the code attempts to retrieve the codelist. If it succeeds, it compares checksums and updates the file on disk if changed. If it fails, then it logs a warning and reads from disk.

When reading/writing to disk, we can use file locking (if the file is shared between workers).

In terms of where to write the file, https://docs.docker.com/storage/tmpfs/ might be appropriate. Alternately, we can write the files to the host (via a mount), so that the on-disk cache survives any worker restarts, and so that Salt can install initial copies of the codelists, to bootstrap this fallback method.

[jpmckinney]

Noting that this does come up in practice.