- Remove `fn_sig()` from Instance.
- Change return value of `AssertMessage::description` to `Cow<>`.
- Add assert to instance `ty()`.
- Generalize uint / int type creation.

…iler-errors

Erase late bound regions from `Instance::fn_sig()` and add a few more details to StableMIR APIs

The Instance `fn_sig()` still included a late bound regions which needed a new compiler function in order to be erased. I've also bundled the following small fixes in this PR, let me know if you want me to isolate any of them.

  - Add missing `CoroutineKind::AsyncGen`.
  - Add optional spread argument to function body which is needed to properly analyze compiler shims.
  - Add a utility method to iterate over all locals together with their declaration.
  - Add a method to get the description of `AssertMessage`*.

* For the last one, we could consider eventually calling the internal `AssertKind::description()` to avoid code duplication. However, we still don't have ways to convert `AssertMessage`, `Operand`, `Place` and others, in order to use that. The other downside of using the internal method is that it will panic for some of the variants.

r ? `@ouz-a`

Opportunistically resolve region var in canonicalizer (instead of resolving root var)

See comment in `compiler/rustc_type_ir/src/infcx.rs`.

The **root** infer region for a given region vid may not actually be nameable from the universe of the original vid. That means that the assertion in the canonicalizer was too strict, since the `EagerResolver` that we use before canonicalizing is doing only as much resolving as it can.

This replaces `resolve_lt_var` and `probe_lt_var` in the `rustc_type_ir` API with `opportunistic_resolve_lt_var`, which acts as you expect it should. I left a FIXME that complains about the inconsistency.

This test is really gnarly, but I have no idea how to minimize it, since it seems to kind of just be coincidental that it triggered this issue. I hope the underlying root cause is easy enough to understand, though.

r? `@lcnr` or `@aliemjay`

Fixes rust-lang#118950