fire debug assertion when expected and found item look the same in diagnostic?

[matthiaskrgr]

Sometimes rustc outputs silly error hints like

   = note: expected trait `for<'a> <for<'a> fn(&'a u32) -> &'a u32 {function2} as FnOnce<(&'a u32,)>>`
              found trait `for<'a> <for<'a> fn(&'a u32) -> &'a u32 {function2} as FnOnce<(&'a u32,)>>`

where "expected" and "found" look like they are the same which is quite unhelpful.
(example from #114883 )

I wonder if we should have a debug assertion in these cases (rustc displaying both types being identical although they shouldn't be) that helps catch these kid of diagnostic mishaps, similar to the "span must not be empty and have no suggestion" and "span must not have overlapping parts" debug assertions.

[fmease]

cc #112985

[matthiaskrgr]

Ouff I did not know there were so many already 😅

[jieyouxu]

@rustbot label +A-lifetimes +D-confusing +T-compiler -needs-triage

[chenyukang]

seems it's fixed for this case:
#114883 (comment)