[red-knot] Narrowing for type(x) is C checks

[sharkdp]

Summary

Add type narrowing for type(x) is C conditions (and else clauses of type(x) is not C conditionals):

if type(x) is A:
    reveal_type(x)  # revealed: A
else:
    reveal_type(x)  # revealed: A | B

closes: #14431, part of: #13694

Test Plan

New Markdown-based tests.

[sharkdp]

Both pyright and mypy are less strict here and infer A instead of A | B. This is unsound, as can be demonstrated by returning B() from get_a_or_b(), and observing that this branch is taken.

[codspeed-hq]

CodSpeed Performance Report

Merging #14432 will not alter performance

_{Comparing david/type-x-is-c-narrowing (2e7cd69) with main (3642381)}

Summary

✅ 32 untouched benchmarks

[AlexWaygood]

Nice!

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[sharkdp]

Ah, the performance regression is probably from removing this (now insufficient) check:

        if !left.is_name_expr() && comparators.iter().all(|c| !c.is_name_expr()) {
            // If none of the comparators are name expressions,
            // we have no symbol to narrow down the type of.
            return None;
        }

I'll look into it.

Edit: fixed now by introducing a similar check.