[llvm] [AMDGPU] Switch V_CNDMASK operands to shrink it into VOP2 (PR #135162)

[Mirko Brkušanin via llvm-commits]

================
@@ -0,0 +1,121 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
----------------
mbrkusanin wrote:

Tests for si-shrink-instructions are usually called shrink-*.ll.
Maybe call this shrink-cndmask.ll

https://github.com/llvm/llvm-project/pull/135162