[llvm] [Docs][AMDGPU] Consistent use of "wave-uniform", "per-lane" and "immediate" (PR #209974)

[Sameer Sahasrabuddhe via llvm-commits]

================
@@ -143,4 +143,4 @@ zero-initialized.
 Despite the absence of ``.async`` in their names, these intrinsics are
 asynchronous.
 
-All arguments must be wave-uniform.
+All arguments must be uniform.
----------------
ssahasra wrote:

Fixed. My original argument was that UniformityAnalysis on AMDGPU returns "uniform" for things like `readfirstlane`, without qualifying it with "wave". But it's not very convincing. I realized that documentation must focus on being correct, and should not be influenced by the _implementation_. Maybe when we get serious with other scopes like "workgroup" we will eventually incorporate that into UA.

https://github.com/llvm/llvm-project/pull/209974