fix: clean up Docker sandbox containers on exit

Hey! Thanks so much for making agent-eval and open sourcing it. I've been using it for some of our own internal evals at Sanity and it's been great so far. Since we use Docker to run our evals, I noticed a few things that could be improved for local DX when running evals through Docker.

This PR makes it so that sandbox containers started via Docker are cleaned up (or at least cleanup is attempted) when the CLI exits. I noticed that after running ctrl + c while running evals locally, the sandbox containers were still running in the background. This should fix that, preventing stale Docker containers from piling up.