Add support for cfg_attr-like attributes

[saks]

Hello uniffi team!

I came across issue #2000 in my code and came up with what may solve this problem for a number scenarios. I do realize that this is not 100% bulletproof, but may work good enough before we figure out a better approach. Happy to add support for more scenarios. Any feedback is welcome.

[bendk]

IIUC, this change makes it so that if a UniFFI attribute appears inside a nested meta list then we apply it to the current item. I think this is what #2086 was doing as well. I'm not sure if it will be the final solution, but it could be a nice hack that solves 80% of the cases. I'm interested to hear what @rafaelbeckel and @jplatte think though.

[bendk]

The code looks correct to me, @rafaelbeckel likes it. Can we merge this one?

[jplatte]

I'd still much prefer the main attribute be constrained to be exactly cfg_attr, and the first argument be constrained to exactly uniffi.

IMHO it is more important that accidentally doing the wrong thing is prevented (i.e. a user writes #[cfg_attr(actually_meaningful_cfg_that_only_applies_sometimes, uniffi::method(name = "conditionally_different_name"))], it appears to work, later they discover that the rename is actually always applied), than supporting the obvious thing (#[cfg_attr(feature = "uniffi", uniffi::attribute(args))]). And even if the obvious thing is changed to work later using an implementation approach like what @bendk suggested in the issue where cfg's are actually properly evaluated, there could still be value in the uniffi pseudo-cfg being supported (since it is shorter to write / easier to read).

[rafaelbeckel]

I have no strong opinion about this, and restricting it to cfg_attr does the job for me just fine.

[saks]

Thank you @bendk, @jplatte and @rafaelbeckel for the feedback. I've limited the new functionality to only support cfg_attr attributes. Please have a look at the updated code and let me know if this matches your expectations.

[rafaelbeckel]

Looks good to me. Limiting it to cfg_attr solves the original issue just fine. I can't think of other use cases. If there is one, someone else will open another issue in the future.

[mhammond]

Can we please add a section at the end of here (I guess?) and maybe even a changelog? :)

[saks]

thank you @mhammond, should have done it earlier. Please see if update to docs is making sense. Copywriting is not necessarily my strength :)

[mhammond]

Thanks!

[bendk]

Looks good to me, thanks for adding those docs.