feat(tests): add nightly workflow

[j4n]

.github/workflows/nightly.yml:

• Scheduled at 02:17 UTC daily (also workflow_dispatch for manual runs).
• based on reusable workflow

Adds .github/workflows/nightly.yml, a daily full-stack integration run of lxc-test.yml workflow across a larger matrix of relay deployments.

• Scheduled 02:17 UTC daily, plus workflow_dispatch for manual runs.
• Exercises cmdeploy (dual-stack DNS and --type ipv4), madmail at --source @latest, the Docker driver against ghcr:main, and the cross-relay test-mini matrix in both directions.
• Based on the @latest resolution built into chore(madmail): port to v2 #31.

Before merge, drop the two blocks marked TEMPORARY in the workflow: the push: trigger on this branch, and cmlxc_version: ${{ github.sha }}, which pins the reusable workflow to the branch under test rather than main.

[missytake]

Looks good to me, but I guess others should look at this as well :)

I think there was also the idea to run https://github.com/chatmail/core/blob/main/.github/workflows/ci.yml against the four different deployment options? afaict you can only pass one CHATMAIL_DOMAIN to it, so it doesn't seem to be cross-server. Which should be enough, too :D

but maybe a cmlxc core-test command is out of scope for this PR, this is already a good start.

How do we get notified on failure?

[hpk42]

it's a bit weird this is living in cmlxc but i guess it's worth a try if this nightly approach reliably works.

[j4n]

it's a bit weird this is living in cmlxc but i guess it's worth a try if this nightly approach reliably works.

I think we could just move it over to the relay repo and call the reusable workflow from there. The question about notification on test failures is relevant: It seems one way to do it would be to allow the action to create an issue when the build fails (and potentially auto-close it when it succeeds again). Or somehow patching cmsend in, with its state in a secret or something.