Remove obsolete Literal deduplication assertion

[AlexWaygood]

Fixes #784.

Summary

Remove the assertion that Literal[[], []].__args__ is ([], []). python/cpython#153914 deliberately changed typing.Literal to deduplicate unhashable arguments, so Python 3.14.7 now returns ([],) instead. On supported Python versions where typing_extensions.Literal aliases typing.Literal, the old assertion therefore contradicts the standard library behavior.

This PR just deletes the assertion entirely: the behaviour here is now version-dependent, and the assertion was always testing an implementation detail that wasn't really that important.

Historical context

• The assertion entered typing_extensions in Fix various things with Literal #145. That PR addressed (🐞) Literal doesn't flatten it's arguments #113 and backported bpo-42345: Fix three issues with typing.Literal parameters cpython#23294 and bpo-42345: Fix hash implementation of typing.Literal cpython#23383. The assertion was copied verbatim with the rest of CPython's LiteralTests.test_args in the first commit of that PR. There was no discussion about this particular assertion on Fix various things with Literal #145.
• The original CPython behavior was intentional. In the original review discussion, a reviewer proposed deduplicating unhashable arguments as well. The PR author noted that changing the shared deduplication helper would slow down Union, and Guido asked whether mutable Literal arguments were actually valid.
• PEP 586 rejects mutable Literal arguments at type-checking time but intentionally does not validate them at runtime. The CPython reviewers therefore agreed not to deduplicate mutable arguments. Guido subsequently requested explicit __args__ coverage, and commit f653d680 introduced the exact assertion that Fix various things with Literal #145 later copied.
• gh-153896: Deduplicate unhashable Literal params cpython#153914 revisited that historical decision and now deduplicates unhashable arguments too; the change was backported to Python 3.13, 3.14, and 3.15. The old assertion is consequently obsolete and does not represent a typing_extensions implementation bug.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.26%. Comparing base (83400e9) to head (6e18c1c).

@@            Coverage Diff             @@
##             main     #785      +/-   ##
==========================================
- Coverage   97.26%   97.26%   -0.01%     
==========================================
  Files           3        3              
  Lines        7860     7859       -1     
==========================================
- Hits         7645     7644       -1     
  Misses        215      215              

Flag	Coverage Δ
3.10	88.85% <ø> (-0.01%)	⬇️
3.10.4	88.85% <ø> (-0.01%)	⬇️
3.11	88.02% <ø> (-0.01%)	⬇️
3.11.0	87.27% <ø> (-0.01%)	⬇️
3.12	87.96% <ø> (-0.01%)	⬇️
3.12.0	87.94% <ø> (-0.01%)	⬇️
3.13	83.39% <ø> (-0.01%)	⬇️
3.13.0	84.09% <ø> (-0.01%)	⬇️
3.14	79.53% <ø> (-0.01%)	⬇️
3.9	89.53% <ø> (-0.01%)	⬇️
3.9.12	89.53% <ø> (-0.01%)	⬇️
pypy3.10	88.68% <ø> (-0.01%)	⬇️
pypy3.11	87.88% <ø> (-0.01%)	⬇️
pypy3.9	89.37% <ø> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/test_typing_extensions.py	98.23% <ø> (-0.01%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[srittau]

Makes sense to me.