Enhance/New Lint: `option_and_then_some` for the `Result` equivalent

[flip1995]

It does not (yet) report anything for the Result equivalent:

fn main() {
    let a = Err::<(), _>("error");
    let b = a.or_else(|_| Err("another error"));
    println!("{:?}", b);
}

Originally posted by @shepmaster in #1459 (comment)

[shepmaster]

Maybe also for Result::and_then (as mentioned by the original suggester, @drystone):

fn main() {
    let a = Ok::<_, ()>(42);
    let b = a.and_then(|_| Ok(22));
    println!("{:?}", b);
}

[shepmaster]

In the first case, I'd expect suggestion to use map_err, the second would be map.

[alex-700]

Hi! If you don't mind, I'll add this to the weekends. Is it better to introduce new lint for it or append to the current (and rename it)?

[flip1995]

You can introduce a new lint for this result_and_then_ok. Maybe also result_or_else_err, but I'm not sure if we should split those up that far.

If you can come up with a good name, for the combination of the three cases, go for it.

[alex-700]

I decided to significantly improve the lint option_and_then_some before generalization to Result cases. More specifically I want to get this lint not only on

let _ = x.and_then(|o| Some(o + 1));

but on

let _ = x.and_then(|o| {
    do_smth(o);
    Some(o + 1)
})

and even on

let _ = opt().and_then(|s| {
    if s.len() < 2 {
        return Some(20);
    } else if s.len() > 20 {
        return Some(s.len());
    }
    Some(30)
});

And I did it. And waste a lot of time to understand that multisuggestion doesn't work (rust-lang/rust#53934).
And now I have a question: is it ok to drop rust-fix ui tests (comment with FIXME after 53934)? Or autofix functionality (only for |s| Some(s)) is more important than possibility to find more interesting cases?

Also I want to unite Option and Result lints into one. What do you think about lazy_logical_instead_of_map name? Or even more "monadic": bind_instead_of_map?

[flip1995]

is it ok to drop rust-fix ui tests (comment with FIXME after 53934)?

Maybe use db.span_suggestion for the simple case and only use multipart_sugg for the other case. But I guess you already did this. That way the simple case is still auto fixable. For testing: just split up the tests in two files: <lint_name>.rs and <lint_name>_multipart.rs, where only on the first file rustfix is run.

As for the lint name: I like the second one better, so I suggest you go with that for now, and we can revisit the naming just before merging the PR (renaming should not be that much work anyway).