all that stuff twice. Also, code reads so much nicer.

1. Produced more unique types than is necessary. This increases memory consumption.
2. Linking the type parameter to its definition *seems* like a good idea, but it
   encourages reliance on the bounds listing.
3. It made pretty-printing harder and in particular was causing bad error messages
   when errors occurred before the `TypeParameterDef` entries were fully stored.

Rename `Polytype` to `TypeScheme` to differentiate type schemes (early bound) from higher-ranked things (late-bound), which also use the `Poly` prefix.

This makes sty only 32 bytes on machines with 64-bit pointers.

This cuts the ty_bare_fn variant to 48 bytes rather than 56. There
doesn't seem to be a noticable memory usage decrease from this.

This current inflates memory use more than 3 times.

Fixes #19707.

Includes a bit of refactoring to store `?` unbounds as bounds with a modifier, rather than in their own world, in the AST at least.

There ain’t an RFC 248, while 438 looks to be what is being referred to:
https://github.com/rust-lang/rfcs/blob/master/text/0438-precedence-of-plus.md

type of a fn item or a fn pointer, which are now differentiated.
Introduce coercion from fn item to fn pointer.

Implement support in the parser for generalized where clauses,
as well as the conversion of ast::WherePredicates to
ty::Predicate in `collect.rs`.

Centralize on using `Binder` to introduce new binding levels, rather than having FnSig carry an implicit binding level.  This means that we be more typesafe in general, since things that instantiate bound regions can drop the Binder to reflect that.

Closes #18433

language. Recommend `move||` instead.

with the fn traits

in most cases, just the error message changed, but in some cases we
are reporting new errors that OUGHT to have been reported before but
we're overlooked (mostly involving the `'static` bound on `Send`).

Fix two bugs in HRTB: 1. Categorize early-vs-late bindings on impls when constructing    generics, so that we don't add unnecessary region parameters. 2. Correct the DeBruijn indices when substituting the self type    into the method signature.

Previously, the DeBruijn index for the self type was not being
adjusted to account for the fn binder. This mean that when late-bound
regions were instantiated, you sometimes wind up with two distinct
lifetimes.

Fixes #19537.

This means that `Fn(&A) -> (&B, &C)` is equivalent to `for<'a> Fn(&'a A)
-> (&'a B, &'a C)` similar to the lifetime elision of lower-case `fn` in
types and declarations.

Closes #18992.

down on its exports. Remove some dead code that is revealed.