Rollup merge of #114594 - compiler-errors:new-solver-resolve-aliases, r=lcnr

Structurally normalize weak and inherent in new solver

It seems pretty obvious to me that we should be normalizing weak and inherent aliases too, since they can always be normalized. This PR still leaves open the question of what to do with opaques, though 💀

**Also**, we need to structurally resolve the target of a coercion, for the UI test to work.

r? `@lcnr`

compiler/rustc_trait_selection/src/traits/structural_normalize.rs

@@ -22,7 +22,9 @@ impl<'tcx> StructurallyNormalizeExt<'tcx> for At<'_, 'tcx> {
         assert!(!ty.is_ty_var(), "should have resolved vars before calling");
 
         if self.infcx.next_trait_solver() {
-            while let ty::Alias(ty::Projection, projection_ty) = *ty.kind() {
+            while let ty::Alias(ty::Projection | ty::Inherent | ty::Weak, projection_ty) =
+                *ty.kind()
+            {
                 let new_infer_ty = self.infcx.next_ty_var(TypeVariableOrigin {
                     kind: TypeVariableOriginKind::NormalizeProjectionType,
                     span: self.cause.span,