[llvm] Add `llvm.vector.partial.reduce.fadd` intrinsic (PR #159776)

[Sander de Smalen via llvm-commits]

================

----------------
sdesmalen-arm wrote:

Given that `isExtOpcode` only returns `true` for integer extends, the code below is never executed for PARTIAL_REDUCE_FMLA. It also seems that none of the tests fail when I remove the code, so there is missing test coverage for it.

https://github.com/llvm/llvm-project/pull/159776