feat: Add --env KEY=VALUE flag to colab run and colab exec to inject environment variables

[mvanhorn]

Summary

google.colab.userdata.get() throws TimeoutException in CLI mode, so users have no clean way to pass secrets (HF_TOKEN, WANDB_API_KEY, etc.) to scripts run via colab run/colab exec. The reporter lists three options; maintainer teeler (CONTRIBUTOR, commented 2026-06-15) acknowledged it as a known issue and tagged the full userdata-API fix for later work because secrets are a separate backend API.

Changes

Add a repeatable --env KEY=VALUE typer.Option (List[str]) to run_command in run.py and to exec_command in execution.py. Parse each entry into a dict (splitting on the first '='), validate the key, and prepend an os.environ-setup prelude to the executed code.

Fixes #48

[teeler]

Clever, I think this is probably fine. I worry a little about notebook drift - notebooks using the userdata api can't also use env vars and vice versa.

But, if this scratches that itch for cli users then it seems reasonable to me.