Rollup merge of #131202 - Urgau:wide-ptrs-compiler, r=jieyouxu

Use wide pointers consistenly across the compiler

This PR replace every use of "fat pointer" for the more recent "wide pointer" terminology.

Since some time T-lang as preferred the "wide pointer" terminology, as can be seen on [the last RFCs](https://github.com/search?q=repo%3Arust-lang%2Frfcs+%22wide+pointer%22&type=code), on some [lints](https://doc.rust-lang.org/rustc/lints/listing/warn-by-default.html#ambiguous-wide-pointer-comparisons), but also in [the reference](https://doc.rust-lang.org/stable/reference/expressions/operator-expr.html?highlight=wide%20pointer#pointer-to-pointer-cast).

Currently we have a [mix of both](https://github.com/search?q=repo%3Arust-lang%2Frust+%22wide+pointer%22&type=code) (including in error messages), which isn't great, but with this PR no more.

r? `@jieyouxu` (feel free to re-roll)

compiler/rustc_hir_analysis/src/check/check.rs

@@ -1105,7 +1105,7 @@ fn check_simd(tcx: TyCtxt<'_>, sp: Span, def_id: LocalDefId) {
         // Check that we use types valid for use in the lanes of a SIMD "vector register"
         // These are scalar types which directly match a "machine" type
         // Yes: Integers, floats, "thin" pointers
-        // No: char, "fat" pointers, compound types
+        // No: char, "wide" pointers, compound types
         match element_ty.kind() {
             ty::Param(_) => (), // pass struct<T>([T; 4]) through, let monomorphization catch errors
             ty::Int(_) | ty::Uint(_) | ty::Float(_) | ty::RawPtr(_, _) => (), // struct([u8; 4]) is ok