[ty] Preserve recursive aliases when rebuilding intersections

[charliermarsh]

Summary

We avoid panics and hangs when recursive aliases appear in mapped intersections. Both negated and positive aliases retain their identity when the intersection does not need to simplify them. For example, indexing a recursive list containing a complement now yields object:

from ty_extensions import Not

type A = list[Not[A] | A]

def f(x: A):
    reveal_type(x[0])  # object

The same applies when specialization puts a recursive tuple alias into a positive intersection:

from collections.abc import Iterable
from typing import Any, assert_type
from ty_extensions import Intersection

type WithIterable[T] = Intersection[Iterable[Any], T]
type B = tuple[WithIterable[B], ...]

def g(x: B):
    assert_type(x[0], WithIterable[B])

The intersection builder expands set operations for distribution and flattening, then inspects the types behind aliases when simplifying the resulting intersections. This preserves existing rules for object, Never, double negation, truthiness, enums, and gradual generics while avoiding repeated expansion of unchanged recursive aliases. Type comparisons still treat aliased and expanded intersections as equivalent, including those containing gradual types.

The invalid x[] form in the issue now reports a syntax error without panicking.

Closes astral-sh/ty#3195.

[astral-sh-bot]

Typing conformance results

No changes detected ✅

Current numbers

The percentage of diagnostics emitted that were expected errors held steady at 98.24%. The percentage of expected errors that received a diagnostic held steady at 98.24%. The number of fully passing files held steady at 134/146.

[astral-sh-bot]

Memory usage report

Memory usage unchanged ✅

[astral-sh-bot]

ecosystem-analyzer results

No diagnostic changes detected ✅

Full report with detailed diff (timing results)

[codspeed]

Merging this PR will degrade performance by 4.19%

❌ 1 regressed benchmark
✅ 175 untouched benchmarks
⏩ 60 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	ty_micro[gradual_intersection_negation]	1.9 s	2 s	-4.19%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing charlie/ty-3195-recursive-complement (9d7b5bb) with main (c33cdaa)}

Footnotes

1. 60 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩