[red-knot] support deferred evaluation of type expressions

[carljm]

Prototype deferred evaluation of type expressions by deferring evaluation of class bases in a stub file. This allows self-referential class definitions, as occur with the definition of str in typeshed (which inherits Sequence[str]).

[codspeed-hq]

CodSpeed Performance Report

Merging #13131 will not alter performance

_{Comparing lazy-bases (b9c1dd3) with main (c6023c0)}

Summary

✅ 32 untouched benchmarks

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[AlexWaygood]

Overall this looks good!

[carljm]

I looked into the regression here, and it seems like it's increased Salsa overhead. This is consistent with the regression only showing up in the incremental benchmark, not the cold benchmark. I guess this is because we now double the number of tracked function memos for definitions, which are our highest-cardinality tracked struct.

[carljm]

I added tracking of whether a given definition has any deferred expressions, which allows us to avoid ever calling infer_deferred_types query on a definition that doesn't have any. Locally this seems to make the regression go away, will see if CodSpeed agrees.

[carljm]

Yay, CodSpeed agrees!