Address repo size

[jackgerrits]

A fresh clone of the repo is currently ~170MiB, and the test dir is ~17MiB. The .git folder is ~128MiB. So there's some decently large files spread through the git history.

One approach I've seen used elsewhere is to place datasets into an archive that is uploaded to blob storage. The repo can then download and uncompress the artifact if they are needed for testing, with the added benefit that if you don't need to test the code you can avoid that cose.

[jackgerrits]

https://stackoverflow.com/a/685422