Investigate list parsing for possible performance improvement

[dhruvmanila]

It seems like there could be performance improvements in list parsing. These methods are used to parse a sequence of nodes, optionally with a comma, which performs error recovery. This is only the case in the new parser.

This PR (#10786) showed a regression which got resolved (at least locally) by removing list parsing in the fast path (single assignment target).

[MichaReiser]

I wonder if that was simply because list parsing might not get inlined (our parser is now so fast that function calls could be expensive :D)

[dhruvmanila]

I also noticed that parse_conditionl_expression_or_higher is not getting inlined anymore after #10809. It could be because there's a new branch in the function:

ruff/crates/ruff_python_parser/src/parser/expression.rs

Lines 211 to 212 in 22d70b0

	if self.at(TokenKind::Lambda) {
	Expr::Lambda(self.parse_lambda_expr()).into()

We could create a new parse_lambda_expression_or_higher which calls into parse_conditional_expression_or_higher and update all references of the latter to use the former:

pub(super) fn parse_lambda_expression_or_higher(&mut self, allow_starred_expression: AllowStarredExpression) -> ParsedExpr {
	if self.at(TokenKind::Lambda) {
		Expr::Lambda(self.parse_lambda_expr()).into()
	} else {
		self.parse_conditional_expression_or_higher(allow_starred_expression)
	}
}

[MichaReiser]

Yeah, or try to force inline by using #[inline]

[dhruvmanila]

Yeah, or try to force inline by using #[inline]

I tried that, it didn't work