[clang] [serialization] No transitive type change (PR #92511)

[Ilya Biryukov via cfe-commits]

https://github.com/ilya-biryukov commented:

Thanks! With all the explanation, the patch makes a lot of sense.
Generally LG, I just have a few more NITs and feel it's good to go.

I am surprised to hear we have less TypeIDs than DeclIDs. My mental model was that for every type Decl we introduce, we'd have at least a few types with it (the type itself, pointer to that type, reference to that type, functions returning that type, etc). And every non-type declaration will have it's own type (the functions often have unique signatures).
Do we end up reusing the types from the current compilation and other modules we imported often enough that the amount of newly introduced types ends up being small? Or is my logic flawed in some ways?

https://github.com/llvm/llvm-project/pull/92511