[llvm] [X86][ARM][AArch64] shouldFoldMaskToVariableShiftPair should be true for scalars up to the biggest legal type (PR #156886)

[via llvm-commits]

AZero13 wrote:

Authors Note:

```armasm
  ldr r0, [r0]
  adds r1, r2, r1
  rors r0, r0, r1
  movs r1, 32
  subs r1, r1, r2
  lsrs r0, r0, r2
  bx lr```
  
  Is a better sequence for the extract case in particular, but that will be for another PR.


https://github.com/llvm/llvm-project/pull/156886