[llvm] [AMDGPU] Select v_lshl_add_u32 instead of v_mul_lo_u32 by constant (PR #71035)

[via llvm-commits]

sstipanovic wrote:

> You're probably better off doing this in a selection pattern. The 24-bit cases are better off with the multiply, and SALU vs. VALU matters too

Switched to selection pattern instead.

https://github.com/llvm/llvm-project/pull/71035