fix: fix type of StatusBase subclasses by calling StatusBase.register in __init_subclass__

[james-garner-canonical]

This is a simple alternative to #1382, following @dimaqq's suggestion to just not use StatusBase.register as a decorator. In this PR we instead do so via __init_subclass__.

[tonyandrewmeyer]

I like it - I think there was a clear expectation that every subclass would register so doing it automatically, while a behaviour change, is ok.

I couldn't find any charms with (StatusBase, which isn't a perfect search for subclassing, but I think it's a good sign that no-one is doing that.

I like that register remains unchanged - we just ignore the returned class. So no need for deprecation warnings etc, and if people are registering different types of class (no idea why) then that would work. And we already allowed registering twice, so anyone doing this:

@Statusbase.register
class OtherStatus(StatusBase):
    name = "other"

Is still fine.

[james-garner-canonical]

Thanks, @tonyandrewmeyer. I also tried ag 'class .*\(.*StatusBase' to see if anyone was subclassing StatusBase and didn't find anything.

If we wanted to deprecate register, we could move the functionality to _register, call that in __init_subclass__ and register, and add a deprecation warning to register. If there's interest in doing that, I'm happy to add it to the PR.

[benhoyt]

This approach is nice and simple -- looks good to me!