update `#[derive(Eq)]` to not need `assert_receiver_is_total_eq`

[cyrgani]

#[derive(Eq)] currently uses the hidden method Eq::assert_receiver_is_total_eq to emit checks that every field of the type being derived on also implements Eq. It would be nice to be able to get rid of it entirely, which requires changing this derive macro to no longer place its checks in this method.

Some implementation notes from #149978 (comment):

---

For this struct:

struct Point {
    x: u32,
    y: u32,
}

Instead of generating this:

#[automatically_derived]
impl ::core::cmp::Eq for Point {
    #[inline]
    #[doc(hidden)]
    #[coverage(off)]
    #[allow(internal_eq_trait_method_impls)]
    fn assert_receiver_is_total_eq(&self) -> () {
        let _: ::core::cmp::AssertParamIsEq<u32>;
    }
}

You would generate this:

#[automatically_derived]
impl ::core::cmp::Eq for Point {}
const _: () = {
    #[doc(hidden)]
    #[coverage(off)]
    fn assert_receiver_is_total_eq() {
        let _: ::core::cmp::AssertParamIsEq<u32>;
    }
};

One problem is that the const should probably also have #[automatically_derived] put on it, but that's currently not permitted.

---

#149978 (comment) contains some more notes on this.

Tracking issue: #152336

[juntyr]

It would probably be good to ensure that the span of the generated const block still points at the derive(Eq)

[KiChjang]

@rustbot claim

[KiChjang]

Probably best if I wait until #144978 is merged, however I already have a working implementation locally, and it generates the following:

#[automatically_derived]
impl ::core::cmp::Eq for Point {}
const _: () = {
    #[doc(hidden)]
    #[coverage(off)]
    const fn assert_receiver_is_total_eq() {
        let _: ::core::cmp::AssertParamIsEq<u32>;
    }
};

The function is const simply because of some types using #[derive_const], and since we're doing the assertion during compilation time anyway, I think it's reasonable to just declare the dummy function as const.

Unsure if we still want to go ahead and allow #[automatically_derived] to be put on const items, but it's a simple switch and if that's desired, I can go ahead with it.