[clang] [Clang] Correctly propagate type aliases' unexpanded flags up to lambda (PR #122875)

[Younan Zhang via cfe-commits]

zyn0217 wrote:

> There is something generally wrong with type aliases which don't use all of their parameters, and this needs more thought.

The current approach - bubbling up the containsUnexpanded flag within a transform - is admittedly not ideal. However, it's the only way to make Clang correctly handle nested packs, within lambdas.

I remembered @cor3ntin once told me Richard had an idea of completely refactoring the pack model as in https://github.com/llvm/llvm-project/issues/9395#issuecomment-980892278. If implemented, we could get rid of this 'workaround'. Maybe we could explore that (probably starting from lambdas?) in the future.

https://github.com/llvm/llvm-project/pull/122875