[clang] [clang] Implement CTAD for type alias template. (PR #77890)

[via cfe-commits]

cor3ntin wrote:

@hokein @sam-mccall to repeat what was said in the language group last week.

We do not think a feature flag is a good fit. 

 - Until clang 19 ships (in 6 months) we do not need to stabilize the feature (but ofc we should avoid regressions, which we would not find if hidden behind a flag), and the flag does not guarantee existing features are not impacted.
 
And if we break trunk, we revert, you get extra test cases... it's perfectly fine

 - The feature should be behind a -std=C++20 flag.
 - We should not update __cpp_deduction_guides until the feature is battle tested

We do that for about every feature that is not concept/module scale (and then again, these things were tses), and this isn't big enough to warrant a novel approach.
  
I still think we should land this soon and we can start a review whenever you are ready.
If you want to land it in an incomplete state, we just need to make sure we have a good collective understanding of the remaining work that would need to be done to bring the feature to completion (tests with fix me, issues, etc)

Does that make sense to you?





https://github.com/llvm/llvm-project/pull/77890