Suggest `Reverse` instead of casting number to signed counterpart and making it negative

What it does

I saw this real-world example today:

multi_labels.sort_by_key(|m| -(m.span.len() as isize));

where len() returns a usize.

My suggestion would resolve it to this:

multi_labels.sort_by_key(|m| Reverse(m.span.len()))

Lint Name

negative_sort_by_key

Category

correctness, suspicious, style

Advantage

• Removes uncessary cast
• Removes panicable cast
• Makes expression more clear

Drawbacks

Importing the Reverse struct is necessary.

Example

use std::cmp::Reverse;

#[derive(Debug, Clone, Copy, PartialEq)]
struct S(usize);

fn sort_a(mut v: Vec<S>) -> Vec<S> {
    v.sort_by_key(|m| -(m.0 as isize));
    v
}

fn sort_b(mut v: Vec<S>) -> Vec<S> {
    v.sort_by_key(|m| Reverse(m.0));
    v
}

fn main() {
    let vec = vec![S(3), S(0), S(0), S(5), S(13), S(18)];
    assert_eq!(sort_a(vec.clone()), sort_b(vec));
}

Start by locating the negative_sort_by_key lint and reviewing the sort_a and sort_b examples in this issue. Confirm how the suggested Reverse form should be recognized and what existing lint tests cover; done means the lint recommends Reverse for this cast-and-negation pattern without changing unrelated cases.