Add :only: option for code cells

[tuncbkose]

I made this primarily for myself but thought should make a PR if there would be any interest in merging. I'm happy to add tests/documentation, but I'd like to get an OK that the implementation is fine first.

Also, as far as I can tell, NbParserConfig.builder_name isn't used by anything, so I just co-opted it.

[bsipocz]

I'm not sure yet if this will be widely used for notebooks, but cannot hurt the addition. Pending that tests and documentation will be added to this PR.

[tuncbkose]

While adding the test, I added a get_latex function to access the built file. For some reason though the built latex file is named python.tex instead of {filename}.tex. I could not figure out why, so if anyone has an idea I'd be interested to know.

[tuncbkose]

There are two errors, 1 is a typo I made and the other is about that Tex file naming issue as far as I can tell. Will take a closer look later.

[agoose77]

I haven't had a long think about this PR, but my initial sense is one of reluctance to merge it. The code-cell directive in MyST-NB is part of an (implicit) standard, and any new features should be something we carefully introduce. MyST-MD is an example of another project that implements the MyST spec.

I wonder here whether a post-process tool would be better suited, e.g. https://github.com/agoose77/sphinx-builder-classes. This extension allows you to add classes to your directives that are ignored for particular builders.

[tuncbkose]

I understand if there may be good reasons not to merge this, but as far as I understand :load: is also not a part of the spec, and it has been sitting here for a while.

About the link, post-processing doesn't always help when you need the code outputs to be dependent on the builder. I appreciate that this is not a straightforward thing to implement in the best way, but maybe something flagged as experimental may be ok, following :load:s example. I'll be using it myself either way.