Fix: application timeout in fast-track if only 1 gpu

[S1ro1]

Description

Small fix, PR #144 introduced a subtle bug that if only 1 gpu was available, it would not call gpu_view, resulting in the interaction defer not being called. As only 1 command calls this view, I moved the defer into the command as it's not view specific.

Checklist

Before submitting this PR, ensure the following steps have been completed:

• Run the slash command /verifyruns on your own server.
  • Run the cluster bot on your server:

    python discord-bot.py

  • Start training runs with the slash command /verifyruns.
  • Verify that the bot eventually responds with:

    ✅ All runs completed successfully!
    

    (It may take a few minutes for all runs to finish. In particular, the GitHub
    runs may take a little longer. The Modal run is typically quick.)
    For more information on running a cluster bot on your own server, see
    README.md.