[llvm] [ExpandLargeFpConvert] Fix bug in int-to-fp expansion. (PR #85370)

[Jay Foad via llvm-commits]

jayfoad wrote:

This commit causes `test/CodeGen/AMDGPU/itofp.i128.ll` to fail in my Release+Asserts build. I don't understand why it didn't fail in the CI build too.

https://github.com/llvm/llvm-project/pull/85370