Separate clauses and goals from the Predicate type

[oli-obk]

Split up the PredicateKind type into Goal and Clauses similar (or exactly :D) like what chalk does. The TLDR is that ParamEnv only contains a small set of PrecicateKind variants and we want to encode this in the type system by only returning Clauses from it.

This projection has been accepted as an MCP: rust-lang/compiler-team#531

Implementation steps

• [SP] Split out Trait, TypeOutlives, RegionOutlives and Projection into Clause enum: #104846
• [SP] #104911 inferred_outlives_crate returns predicates, but they are always only TypeOutlives or RegionOutlives
  • may kill perf, so let's see if it works to just return ty::Clause where it now returns ty::Predicate and build the predicates at the caller site of inferred_outlives_of
• doing the same thing for explicit_predicates_of is gonna be a bit more involved
• Improve perf of #104846
• start with gather_explicit_predicates_of and convert at the caller sites
• Bounds::predicates should return an iterator over Clauses (Make Bound::predicates use Clause #112734)

Follow-up work after #112938:

• Investigate making TypeFoldable<TyCtxt<'tcx>> for ty::Clause<'tcx> implementation less weird:

  rust/compiler/rustc_middle/src/ty/structural_impls.rs

  Line 672 in 2efe091

  .expect("no sensible folder would do this"))

• Clean up the elaborator since it should only be emitting child clauses, not predicates
• Rename identifiers like pred and predicates to clause if they're actually clauses around the codebase
• Validate that all of the ToPredicate impls are acutally still needed, or prune them if they're not

leave a comment when you start working on one of these steps so that no one else duplicates work.

[obeis]

@rustbot claim

[spastorino]

@oli-obk thanks for opening this, I'd be happy to review and/or help/mentor if needed.

@obeis feel free to reach out on Zulip.

[GentBinaku]

@rustbot claim

[BoxyUwU]

PredicateKind::ConstEvaluatable and PredicateKind::WellFormed can all appear in the param_env/predicates_of so should be moved to Clause as well (PredicateKind::ConstEquate might too although we might also end up removing that so 🤷‍♀️)

[spastorino]

@GentBinaku are you still actively working on this?

[dswij]

@rustbot claim

[zirconium-n]

@rustbot claim

[zirconium-n]

I think I'm a little bit unsure about the task.

Validate that all of the ToPredicate impls are acutally still needed, or prune them if they're not.

What count as an unneeded impl? Some impls can be removed without error, but theres one gray area I'm not sure. That is the impl is not there to satisfy a trait bound but just a helper function. Does this count the impl needed or not? We can just inline the impl body?