Resolve parametrized type aliases before checking

[afonsojanu]

Fixes #579.

A subscripted PEP 695 alias like Boxed[int], for type Boxed[T] = list[T], has Boxed as its origin rather than being a TypeAliasType itself, so the existing type(annotation) in type_alias_types check missed it entirely. The annotation fell through unresolved and nothing checked the value against it, so any value passed silently. That's a worse failure mode than a crash, since nothing tells you the check never ran.

Added resolve_type_alias(), which handles both the bare alias case that already worked and the parametrized one, substituting the alias's type arguments into its __value__ the same way subscripting list[T] with int would. Used it at both call sites that were special-casing bare aliases before, check_class and check_type_internal.

Extended dummymodule_py312.py with a function using a parametrized alias and added matching success/failure tests to test_instrumentation.py's TestTypeAlias class, covering both the typechecked and importhook instrumentation paths. Confirmed the new failure test fails against the unpatched code (with a confusing, unrelated error about the return value rather than a clean rejection of the argument) and passes with the fix.

Ran the full suite before and after: same 9 pre-existing failures both times, all environment-only (subprocess tests that need the package actually installed, an unregistered pytest-plugin option, a local mypy version mismatch) and unrelated to this change. Everything else green, 530 passing.

[coveralls]

coverage: 94.873% (+0.02%) from 94.853% — afonsojanu:fix/parametrized-type-alias-skips-check into agronholm:master