[llvm] [AArch64] Add support for 16/32/64-bit floating-point atomic read-modify-write ops (PR #125686)

[Jonathan Thackray via llvm-commits]

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

----------------
jthackray wrote:

Sure. I can modify the `generate-tests.py` script to test with both `+lsfe` and with `+nolsfe`. This will also affect all the other tests, since these will test with `+noxxx` for each feature.

https://github.com/llvm/llvm-project/pull/125686