Expose an API for polymorphic drop glue translation

[Nadrieril]

At my $dayjob I work on a rustc driver called Charon, whose main selling point is its ability to accurately represent polymorphic Rust code. In adding support for drops I ran into an issue: the code that generates drop glue shim expects monomorphized types and ICEs when used in some polymorphic contexts. Example output:

error: internal compiler error: compiler/rustc_middle/src/ty/sty.rs:373:13: cannot find `K/#0` in param-env: ParamEnv {
    caller_bounds: [
        Binder { value: HostEffectPredicate { trait_ref: <T as std::marker::Destruct>, constness: Maybe }, bound_vars: [] },
        Binder { value: TraitPredicate(<T as std::marker::Destruct>, polarity:Positive), bound_vars: [] },
    ],
}

thread 'rustc' panicked at compiler/rustc_middle/src/ty/sty.rs:373:13:

I recall a Zulip discussion about this but cannot find it.

This PR exposes build_drop_shim and gives it a TypingEnv argument, for use in rustc drivers like mine.

Disclosure: I used LLM assistance to make this very complicated PR. This consisted in me repeatedly telling it to do something simpler until I eventually cleaned it up myself. I like the test it wrote tho.

[rustbot]

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

[rustbot]

r? @JonathanBrouwer

rustbot has assigned @JonathanBrouwer.
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 73 candidates
• Random selection from 16 candidates

[Nadrieril]

Wait why on earth would this be backend-dependent 🤔

[Nadrieril]

Investigation reveals some rather broken state of affairs: calling tcx.fn_sig on the DefId for the DropGlue lang item is what ICEs, and it looks like some mixup between DefIds. So I've just ignored the gcc backend.

[Kobzol]

I'm not sure if these situations are interesting for cg_gcc maintainers, but just to make sure - CC @rust-lang/wg-gcc-backend