Test a new setup to deploy to Cachix even for PR from forks.

[Zimmi48]

The explanation can be found in the commit pushed to the pull-request-target branch:

Use the pull_request_target event instead of the pull_request event to deploy to have access to secret environment variables in all cases. We now skip the actions/checkout@v2 step because it wouldn't check out the correct commit. This is safe (no risk of linking the secret variables) even with untrusted code because nix-build will build it in an isolated environment.

I'm opening this fake PR to test the behavior it gives for a PR from a fork (does it report a CI status like it would for a pull_request event? EDIT: yes, it does, with a slightly different name).

[Zimmi48]

Closing / reopening to hopefully trigger a workflow using the updated version in the base ref (that should now test the automatically created merge commit instead of the head of the PR).

[Zimmi48]

Testing was conclusive. This PR can be closed.

[Zimmi48]

Testing one last configuration.

[palmskog]

@Zimmi48 should we keep the pull-request-target branch around? Or can we delete it?

[Zimmi48]

Let's delete it until the corresponding PR on the templates repository is updated (which I won't do before the corresponding PR on the Coq Nix Toolbox repository is merged).

[Zimmi48]

Tested to work when PR has conflicts (tests pull/XX/head) and no conflicts (tests pull/XX/merge).