fix: no suggest ref match when expected generic ref

[A4-Tacks]

Example

fn foo<T>(s: &T) {}
fn main() {
    let ssss = &mut 2i32;
    foo($0);
}

Before this PR

lc ssss &mut i32 [type_could_unify+local]
lc &ssss [type+local]
md core::  []
fn foo(…) fn(&T) []
fn &foo(…) [type]
fn main() fn() []
fn &main() [type]

After this PR

lc ssss &mut i32 [type_could_unify+local]
md core::  []
fn foo(…) fn(&T) []
fn &foo(…) [type]
fn main() fn() []
fn &main() [type]

[rustbot]

This PR was rebased onto a different master commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[ChayimFriedman2]

Worth thinking whether we should check coercion instead, which will also fix that.

Also, we might want to omit the &foo() and &main() suggestions, given that they return () (and in general any &() suggestion)

View changes since this review