Upload cookbook docs to artifacts

[Ifropc]

What

Uploads cookbook docs to artifacts

Why

To decouple cookbook docs from our actual release and avoid potential breakage
docs PR: stellar/stellar-docs#1439

Known limitations

I manually uploaded artifact produced by the latest run to 22.6.0 release

[fnando]

Great idea!

[leighmcculloch]

The cookbook and full help are different things. The full help docs is referred to as the user manual in the docs, so I think we should rename the job to User Manual. We could rename the .md file to USER_MANUAL too maybe for consistency.

[Ifropc]

Do you think we should upload FULL_HELP_DOCS.md and cookbook.tar.gz as separate artifacts?
Because the artifacts are not intended to be used by the end user, just by our docs, so I think it makes sense to keep it in 1 tar, and keep the naming the same to avoid confusion by developers

[leighmcculloch]

I think they should be separate. And named by what they contain. The user manual is at least maybe interesting to others.

[leighmcculloch]

To decouple cookbook docs from our actual release and avoid potential breakage

The PR description says the aim is to decouple cookbook docs from the CLI release, however this PR seems to explicitly make the cookbooks a release artifact, therefore further coupling the cookbooks to the release. 🤔

Is the goal to be able to regenerate and rewrite the release cookbook artifact in cases when the cookbook has an error? 🫨

Setting up a process to rewrite release artifact's is somewhat terrifying, as the release artifacts should be built one time, from the release itself and never tampered with.

The underlying issue seems to me to be that the cookbooks in this repo are not being tested, and if they're to live in this repo instead of the stellar-docs repo, they should probably be tested here. Have we explored that option?

[Ifropc]

Yes, the idea is to combine this approaches (see #1908)
The reason we want to put docs into artifacts is that if somehow we still break the docs, we can simply just replace broken artifact with a working one (though with docs validation this should never happen)

[Ifropc]

We decided not to to with this approach