[llvm] [AArch64][GlobalISel] Take abs scalar codegen closer to SDAG (PR #84886)

[Matt Arsenault via llvm-commits]

arsenm wrote:

> Even if SMAX and lowerAbsToAddXor are legal, shouldn't we have the means to say for FOO cores, we prefer SMAX, and for BAR cores we prefer Xor?

You can always put the decision logic in custom lowering, but I think the default lower should try to guess a reasonable default 

https://github.com/llvm/llvm-project/pull/84886