[red-knot] Add type inference for loop variables inside comprehension scopes

[AlexWaygood]

Summary

This adds type inference for loop variables inside comprehension scopes. The type-inference tests are passing, but it seems to currently trigger a panic in the corpus tests.

Async comprehensions will require different logic, so for now loop variables in these are inferred as Unknown still. In order to distinguish between async comprehension definitions and synchronous comprehensions definitions, a new is_async field is added to ComprehensionDefinitionKind and various other structs for tracking comprehension definitions.

Test Plan

cargo test -p red_knot_python_semantic and cargo test -p red_knot_workspace

[codspeed-hq]

CodSpeed Performance Report

Merging #13251 will not alter performance

_{Comparing alex/comprehension-inference (86aacad) with main (ac720cd)}

Summary

✅ 32 untouched benchmarks

[MichaReiser]

Just FYI: I think the benchmark is panicking now. I see some std::io::Write show up in the benchmarks.

[AlexWaygood]

lots of things are currently broken here 😆 I just figured I'd make a draft PR to show where I'm at (and in case it's obvious to anybody else what the cause of the panics is)

[MichaReiser]

lots of things are currently broken here 😆 I just figured I'd make a draft PR to show where I'm at (and in case it's obvious to anybody else what the cause of the panics is)

Haha, okay. I was shocked by the performance regression, so I had to take a quick look at why it was that bad. I was relieved to see that it was probably caused by the benchmark crashing.

[AlexWaygood]

I was shocked by the performance regression, so I had to take a quick look at why it was that bad. I was relieved to see that it was probably caused by the benchmark crashing.

Oh the PR also currently adds a dbg!() call in quite a hot location so that I can see what's going on more clearly. I assume that doesn't help 😄

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[AlexWaygood]

Haha, okay. I was shocked by the performance regression, so I had to take a quick look at why it was that bad. I was relieved to see that it was probably caused by the benchmark crashing.

All gone now 🎉

[carljm]

Looks great!!

[carljm]

All of the tests! Looks excellent, merge away

[AlexWaygood]

Thanks very much for the help debugging this!