fix: respect transform output type for __in lookups (e.g. __year__in)

[mvanhorn]

I have made things!

Model.objects.filter(start_date__year__in=...) (and other transform-plus-__in lookups) now expects the transform's output element type instead of the base field's type, so start_date__year__in=list[int] type-checks and a list[str] is correctly flagged.

Related issues

Fixes #3411

AI Policy

• I have read and agree to the AI Policy, removed any "Co-Authored-By" lines attributing coding agents, and manually reviewed the final result

Description

DjangoContext.resolve_lookup_expected_type wrapped the base field's exact type in Iterable for the In branch, ignoring any preceding transform (regression in 6.0.5, confirmed by @sobolevn and @ngnpope on 2026-06-02). For start_date__year__in it expected Iterable[str | date] instead of Iterable[int]. The fix uses the transform's output type for the __in element type when a transform precedes the lookup, leaving plain lookups unchanged. A test_filter.yml case covers the valid list[int] and invalid list[str] shapes; the full test_filter.yml suite passes (18 passed, 0 failed).

[ngnpope]

Three quick comments:

• Can you extract this transform type lookup to a helper function? I'm sure this will be needed in more places.
• Can you check if anything needs to be done when filtering on an annotated field?
• More tests should be added for other transforms, not just year.