[llvm] [AMDGPU] Fix FMA Combine (PR #119217)

[Jay Foad via llvm-commits]

================
@@ -0,0 +1,70 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
----------------
jayfoad wrote:

Can you find a better place for this? We already have "fma-combine.ll" (with a hypen instead of an underscore). Maybe it could go in `fdot2.ll` which is where tests we added originally for this combine: https://reviews.llvm.org/D49146

https://github.com/llvm/llvm-project/pull/119217