[llvm] fix `llvm.fma.f16` double rounding issue when there is no native support (PR #171904)

[Folkert de Vries via llvm-commits]

folkertdev wrote:

the test failures are kind of annoying

the `Generic` tests sort of require all backends, and I don't really want to build _everything_. The AMDGPU one succeeds locally, but also its name suggests it deals with `bf16`, so maybe it actually should not change at all?


https://github.com/llvm/llvm-project/pull/171904