[llvm] [LLVM] Change Intrinsic::ID to encode target and intrinsic index (PR #113576)

[Rahul Joshi via llvm-commits]

jurahul wrote:

> > Have you checked whether this does help incremental builds in practice?
> 
> I checked and it does seem to help. Adding a new AMDGPU intrinsic did not cause any other target's `.cpp` files to be rebuilt. (But TBH I'm still not a huge fan of this patch, just because it seems like extra complexity that we don't really need.)

Thanks for checking. I'd say if we were actually going forward with compiling out intrinsics for disabled targets, then this mildly extra complexity might have been worth it. But given that that is not going to happen, this reduces to a pure refactor/NFC for better separation between targets. I'll let the relevant folks decide if it's worth committing.  

https://github.com/llvm/llvm-project/pull/113576