Enforce @attr uses to always provide a return type

[lgoettgens]

We have a lot of places in our codebases where people use @attr without specifying a return type, even though in many of these cases they are very easy too deduce (see e.g. all of the dim and is_* cases in oscar-system/Oscar.jl#4059). I don't know if this is due to people not knowing about the possibility, forget it, or just don't care. But at least for me, it is usually the second one (see e.g. the Lie algebra changes in oscar-system/Oscar.jl#4059).

Thus I propose to deprecate (in the next breaking release) the use of 1-arg @attr (with only a method expression) in favor of explicitly requiring people to write @attr Any if they don't want to think of the return type. I think this would be a good measure for enforcing better inferrable code (by just making it harder to shadow inference by accident). And this is something that can be easily adapted by a search-and-replace when doing the AA compat bump downstream.

As always, the change in deprecations.jl does not deprecate anything right now, it just provides Any as the return type to the 2-arg version, and makes it easy to deprecate it in a future breaking version bump.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.03%. Comparing base (75a8537) to head (018f622).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1782      +/-   ##
==========================================
- Coverage   88.03%   88.03%   -0.01%     
==========================================
  Files         117      117              
  Lines       30094    30088       -6     
==========================================
- Hits        26494    26488       -6     
  Misses       3600     3600              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[lgoettgens]

@fingolfin the Singular test failure is unrelated. Can you force-merge this?

[thofma]

Sorry, I am too slow, but I think this is a good idea.