[llvm] [X86][CostModel] Add per-shape gather/scatter cost tables for AMD znver4+ (PR #199488)

[via llvm-commits]

================
@@ -227,6 +227,11 @@ Makes programs 10x faster by doing Special New Thing.
 
 ### Changes to the X86 Backend
 
+* The cost model now uses per-shape masked gather/scatter costs measured on
----------------
MattPD wrote:

The release note and `Non-goals` section now mention the generic remainder-lane change, but they do not yet describe its full cross-target scope.

https://github.com/llvm/llvm-project/pull/199488