Update CI tooling for raku/doc

[melezhik]

Hi ! After a short talk with @coke creating this suggestion ticket.

https://irclogs.raku.org/raku/2023-02-08.html#01:31-0001

[coke]

My cleanup has probably broken the existing CI we had for tests @dontlaugh, any recommendations for how we should be running our CI tests?

We need:

• a prebuilt raku and zef, (that we can update after each release) and prove (or perhaps prove6)
• preferably pre-installed deps
• A way to (manually is fine) rebuild the test agent when deps change, or raku is updated, etc.

In my opinion, we're fine running tests ONLY on the files that changed for the given commit, but if we do that, we need to definitely monitor when they failed and make them green when they're clear. This makes it more complicated than just running all tests for everything and then we can ignore all previous failures when we get any 100% pass.

[dontlaugh]

I have some infra work to do to facilitate this. Here is the architecture I'm going to pursue (it's what I've implemented for $DAYJOB, as well as for my personal servers).

Compute Substrate

• We provision a linux server running LXD.
• We use Packer to create pre-baked LXD images that
  • Can build Raku and zef from source
  • Install CI agents for running builds
  • Can provide zef, raku, etc. to a CI build process at runtime

The current CI setup for Raku/doc-website is effectively doing all this already, but I need to migrate that stuff into the shared Hetzner account where we are now running docs.raku.org

Config Backup

The Packer files will be checked into Raku/infra; container builds (in this case, LXD containers) are not exactly idempotent (since running apt get foo can technically yield different results over time), but it's good enough.

---

Our current CI for Raku/doc-website already uses this architecture. I just need to re-implement it in the common Hetzner account.