trait solver: Track canonical response universe assumptions

[Dnreikronos]

Fixes #159889

Canonical responses recreate universes for placeholders created while evaluating a query. Region constraints are handled inside the query, but opaque type constraints can still carry those placeholders back to the caller. Without an assumptions entry, eager placeholder handling hits the unwrap() in get_placeholder_assumptions.

Record empty assumptions when recreating response universes under -Zassumptions-on-binders. idk if empty assumptions is the final shape irl, but imo keeping this next to create_next_universe is the right local fix. It restores the invariant where the universe enters the caller and matches bookkeeping from other binder entry paths. fyi the issue repro now reports E0277 instead of ICEing.

[rustbot]

r? @nnethercote

rustbot has assigned @nnethercote.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler
• compiler expanded to 74 candidates
• Random selection from 16 candidates

[nnethercote]

r? types

[oli-obk]

r? @lcnr

[lcnr]

r? BoxyUwU

[rustbot]

BoxyUwU is currently at their maximum review capacity.
They may take a while to respond.

[lcnr]

Handling opaques here is interesting :/ I guess this might be fine as a temporary hack, I've been chatting about this with Boxy before and rust-lang/trait-system-refactor-initiative#271 partially removes the need for this, but to fully remove it will take a while probably

[lcnr]

we should at least have a FIXME here I think

[Dnreikronos]

Added the FIXME, fyi. imo this workaround is a bit gross, but I think the damage is fairly limited here because the query already handles the region constraints. The annoying bit is that opaque type constraints can still carry query-created placeholders back out through the canonical response, so the recreated universes still need an assumptions entry.

idk if trying to compute more precise assumptions here is worth doing asap. The irl fix seems closer to the opaque storage work in rust-lang/trait-system-refactor-initiative#271, so I left this as an empty-assumptions temporary hack for now. btw ltm if you want the FIXME to point at that issue directly.