CLN,TYP Remove string return annotations

[MarcoGorelli]

• Ensure all linting tests pass, see here for how to run them

this was just done with

    parser = argparse.ArgumentParser()
    parser.add_argument('paths', nargs='*', type=Path)
    args = parser.parse_args()
    for path in args.paths:
        text = path.read_text()
        text = re.sub(r'-> "(\w+)"', '-> \\1', text)
        path.write_text(text)

There's other to do (which I'll get to in other PRs), but this already gets rid of a big load of them

[simonjayhawkins]

nice. merge on green.

[jbrockmendel]

Not a huge deal, but I like the string versions since they make the dependency structure clearer, i.e. I’m not going to incorrectly infer that Series is in the namespace.

[MarcoGorelli]

Sure, but you can't click and go to a definition if it's a string version (this is my main reason for wanting to make this kind of change)

I think it only makes the dependency structure clearer if string ones are used consistently (currently there's a mix) - up to you anyway, my preference here isn't too strong

[simonjayhawkins]

Sure, but you can't click and go to a definition if it's a string version (this is my main reason for wanting to make this kind of change)

some discussion in #36034 (comment)

I think it only makes the dependency structure clearer if string ones are used consistently (currently there's a mix)

agreed. You can (and people have) quote types even if the type is imported at runtine, or quote the whole type expression such as "List[grouper.Grouping]" which the PR resolves.

"we" agreed (#36034 (comment)) on going this route. Do we want another discussion?

[jbrockmendel]

Do we want another discussion?

Nope, I'm pretty happy with my "defer to Simon when it comes to annotations" policy.

[MarcoGorelli]

OK, merging following #39174 (review) and #39174 (comment)

[simonjayhawkins]

Thanks @MarcoGorelli