Rollup merge of #162262 - sjwang05:forall-binder-ice, r=BoxyUwU

Avoid manually instantiating some binders in error reporting with `-Znext-solver`

If we call `enter_forall` on a binder before passing it to the new solver, the placeholders that come from instantiation get canonicalized before solving happens. This can be a problem if we end up equating one variable whose universe can't name another variable's, since the equate will succeed in the solver since canonicalization sets all universe indices to the root universe but fail when we actually construct the response with fresh vars at the correct universes, which causes an ICE. We now just hand the unopened binder to the trait solver directly, since (I believe) both solvers are able to handle that just fine.

This came up while I was looking at https://rust-lang.zulipchat.com/#narrow/channel/618216-t-types.2Fcall-for-participation/topic/next-solver.20UI.20test.20triage/with/612965995, and the `universe-error-host-effect.rs`/`poly_select` change was basically the same type of ICE that I found while investigating the first one.