Detect overlapping supertype variants via leaf type IDs

[Veykril]

This can cause hard to debug issues and is never useful so we should recognize this. Ideally at compile time but comparing type ids at compile time is not yet possible

[netlify]

👷 Deploy Preview for salsa-rs processing.

Name	Link
🔨 Latest commit	d1c1ee1
🔍 Latest deploy log	https://app.netlify.com/projects/salsa-rs/deploys/69ca68f04786c800089c56cb

[netlify]

✅ Deploy Preview for salsa-rs canceled.

Name	Link
🔨 Latest commit	d1c1ee1
🔍 Latest deploy log	https://app.netlify.com/projects/salsa-rs/deploys/69ca68f04786c800089c56cb

[codspeed-hq]

Merging this PR will not alter performance

✅ 13 untouched benchmarks

---

_{Comparing Veykril:push-wmkonznwmxrt (d1c1ee1) with master (2f687a1)}

[MichaReiser]

Do you know what the reason is for the perf regression?

[Veykril]

did you rerun it just now? EIther way no clue what was up with codespeed there

[MichaReiser]

Yeah, I did re-run it.

I do plan to look at this (unless @ibraheemdev or someone of the r-a team beats me to it). I'm just a bit swamped with code reviews across projects, so it may take me until next week to get to it.

[Veykril]

Yea no rush, this isn't a feature anyways (and r-a right now can't upgrade since we still need to investigate the fallback immediate cycle handling regression)

[MichaReiser]

Could we use https://docs.rs/static_assertions/latest/static_assertions/macro.assert_type_ne_all.html or does it perform structural equality?

[Veykril]

We can't do that unfortuantely, we need to transitively check all contained variants which we can't do at macro expansion time. That is

#[derive(Supertype)]
enum E {
    A(A),
    B(B),
}

#[derive(Supertype)]
enum B {
    A(A)
}

would be needed to be detected here. Supertype effectively flattens this variant tree. Thats what we want to guard against here, as that means that salsa will turn a E(B::A(A)) into an E::A(A), so variants lose their meaning if they overlap