fix(type-analysis): preserve nested Final types in ClassVar

[daleselaji-dev]

Problem

On Python 3.13 and newer, ClassVar[Final[T]] is accepted but the annotated class variable resolves to Any instead of preserving T.

Root Cause

The type analyzer allows nested Final while analyzing ClassVar, but the Final[...] special-form branch still returns an error AnyType after accepting the nested form. This silently discards the inner type.

Solution

When nested Final[...] is allowed and has a type argument, analyze and return that inner type. The existing diagnostic remains unchanged for contexts where nested Final is not permitted.

Changes

• Preserve the inner type for accepted nested Final[...] annotations.
• Add a Python 3.13 regression case asserting that ClassVar[Final[dict[str, int]]] reveals the dictionary type.

Testing

• Baseline on current master (b974556f): the issue reproducer revealed Any under --python-version 3.13.
• py -3.10 -m pytest 'mypy/test/testcheck.py::TypeCheckSuite::check-final.test' -q: 82 passed.
• py -3.10 -m black --check mypy/typeanal.py: passed.
• py -3.10 -m ruff check mypy/typeanal.py: passed.
• py -3.10 -m compileall -q mypy/typeanal.py: passed.
• py -3.10 -m mypy --config-file mypy_self_check.ini -p mypy: 196 source files.
• git diff --check: passed.
• The full repository test suite was not run.

Compatibility/Risk

This changes only the result of an already-accepted nested Final[...] type form. Unsupported nested Final uses continue to report the existing validation error; no runtime behavior or public runtime API changes.

Notes for Reviewer

The regression test uses --python-version 3.13 while running the test harness on Python 3.10, so it exercises the version-gated type-analysis path without requiring a local Python 3.13 interpreter.

Linked Issue

Fixes #21906

[github-actions]

According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅