Account for incompatible universes and higher-ranked subtyping.

The idea here is that an incoming query may refer to some universes,
and they query response may contain fresh universes that go beyond
those. When we instantiate the query response in the caller's scope,
therefore, we map those new universes into fresh universes for the
caller.

Lots of annoying boilerplate.

Instead, we talk about:

- creating the "next" universe
- universes "extending" one another
- and `u1.can_name(u2)`, meaning that `u1` contains all names from `u2`

The only name was silly. U1 can contain everything from U0 *plus* more
things.

This commit introduces language items for `Arc` and `Rc` so that types
can later be checked to be `Arc` or `Rc` in the NLL borrow checker. The
`lang` attribute is currently limited to `stage1` as it requires a
compiler built with knowledge of the expected language items.

This seemed like a good way to kick the tires on the
elided-lifetimes-in-paths lint (#52069)—seems to work! This was also
pretty tedious—it sure would be nice if `cargo fix` worked on this
codebase (#53896)!

During metadata loading, the AdtDefs for every ADT in the universe need
to be loaded (for example, for coherence of builtin traits). For that,
the attributes of the AdtDef need to be loaded too.

The attributes of a struct are duplicated between 2 def ids - the
constructor def-id, and the "type" def id. Loading attributes for both
def-ids, which was done in #53721, slowed the compilation of small
crates by 2-3%. This PR makes sure we only load the attributes for the
"type" def-id, avoiding the slowdown.

This completely splits the IS_NON_EXHAUSTIVE flag. No functional
changes intended.

Structs and enums can both be non-exhaustive, with a very different
meaning. This PR splits `is_non_exhaustive` to 2 separate functions - 1
for structs, and another for enums, and fixes the places that got the
usage confused.

Fixes #53549.