tmt: Try to describe how to reproduce tmt+packages locally

[cgwalters]

The whole chain of packit -> tmt+testing-farm is very complicated and I struggle to wrap my head around it even without involving bootc.

With a bit of AI assistance we came up with this, which I didn't try yet truly running and clearly if we were to do so we'd need to automate it.

And doing that really is probably going to pull us to split out "build an rpm" from "build a container" because the former is what the packit flow wants today.

xref teemtee/tmt#1018

[jeckersb]

I will look at this in a bit, but note that this is relevant to my interests because the functional tests in bodhi are failing because apparently nu isn't installed and I figured it had something to do with the goofy way we provision things in tmt/testing-farm but I haven't dug much more into it.

[cgwalters]

Yes...I think that may be due to packit != dist-git CI...https://src.fedoraproject.org/rpms/bootc/pull-request/59 may be aiming to help this?

[cgwalters]

cc @henrywang

[henrywang]

What about put RPM building commands into justfile? I can work on it.

[cgwalters]

OK sounds good; I looked at this briefly and it was a bit complicated at least to satisfy the constraints I wanted to achieve.

My initial goal was having the default just (i.e. a container build) implicitly do an RPM build but what is an absolute requirement in this is we preserve the fast incremental build flow. One thing I got hung up on here is ensuring that the Rust build still uses the Rust caching.

A constraint I would also like to retain is that the RPM (in general, package) build infra is always itself in a container and isn't required in the outer build env (e.g. GHA runner, dev container).

[henrywang]

If we can build bootc RPM in Containerfile, we can drop Packit build and make localy test easy.