Figure out how to copy data from coordinator to seeders

[gmega]

I started under the assumption that the test runner and the Deluge nodes would be able to share a filesystem (that was the initial plan with Wings), but:

1. this creates performance artifacts, particularly if we're mounting the same persistent for everyone;
2. Hetzner doesn't support ReadWriteMany.

We therefore need to change approach. While I could generate data locally at the clients, this would mean modifying the client images for an experiment, which I don't want to do. I'm therefore going to try to use the Kubernetes Python client and implement a different uploader.

I'll also adjust the PVCs so they use local storage. This should eliminate performance artifacts.

[gmega]

This was fixed by codex-storage/bittorrent-benchmarks#2