Update test expectations for aarch64

[nikic]

vcgez/vclez now generate different (and more obvious) instructions.

fmadd/fmsub are now preferred over the slower fmla/fmls in more cases. I believe this is specific to the LANE=0 case, but I couldn't figure out how to switch testing to use LANE=1 without also breaking the case where only LANE=0 is valid.

[rustbot]

r? @Amanieu

rustbot has assigned @Amanieu.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[nikic]

I believe the fma change is llvm/llvm-project@38c92c1. I didn't try to track down the other one.