[refurb] Implement reimplemented-starmap rule (FURB140)

[SavchenkoValeriy]

Summary

This PR is part of a bigger effort of re-implementing refurb rules #1348. It adds support for FURB140

Test Plan

I included a new test + checked that all other tests pass.

[github-actions]

PR Check Results

Ecosystem

✅ ecosystem check detected no changes.

[MichaReiser]

The use of a generic here means that Rust generates a dedicated implementation for each T instantiation.

I think it may be better to define a AnyComprehension enum (or similar) with a variant for each node that you want to support instead of using a trait. See

ruff/crates/ruff_python_formatter/src/expression/expr_yield.rs

Lines 10 to 35 in c05e462

	pub(super) enum AnyExpressionYield<'a> {
	Yield(&'a ExprYield),
	YieldFrom(&'a ExprYieldFrom),
	}
	impl<'a> AnyExpressionYield<'a> {
	const fn is_yield_from(&self) -> bool {
	matches!(self, AnyExpressionYield::YieldFrom(_))
	}
	fn value(&self) -> Option<&Expr> {
	match self {
	AnyExpressionYield::Yield(yld) => yld.value.as_deref(),
	AnyExpressionYield::YieldFrom(yld) => Some(&yld.value),
	}
	}
	}
	impl Ranged for AnyExpressionYield<'_> {
	fn range(&self) -> TextRange {
	match self {
	AnyExpressionYield::Yield(yld) => yld.range(),
	AnyExpressionYield::YieldFrom(yld) => yld.range(),
	}
	}
	}

[SavchenkoValeriy]

I'm not a huge fan of such a solution. With the trait, we don't need to perform any runtime checks for the given type. With the enum, we do them 4 times: element, generators, range, and try_fix. These seem not critical in any way, but one can argue that bloating from instantiating the same (rather small) generic 3 times is kind of similar.

I guess I can see how an enum solution is better for decoupling, if I remove try_fix from that interface altogether. Then it can be used in other places that share comprehensions. On the other hand, this interface can't include DictComprehension without making element optional and adding more runtime checks.

Sorry, for the rant. If you insist, I can change it, but I hope you see my point here 😄

[MichaReiser]

I have a few coding nits. Leaving it to @charliermarsh to review the rule logic

[codspeed-hq]

CodSpeed Performance Report

Merging #7253 will degrade performances by 2.38%

_{Comparing SavchenkoValeriy:refurb/furb140 (ea69b1c) with main (c6ba7df)}

Summary

❌ 1 regressions
✅ 24 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	main	SavchenkoValeriy:refurb/furb140	Change
❌	linter/all-rules[numpy/ctypeslib.py]	34.2 ms	35.1 ms	-2.38%