[llvm] [VectorCombine] Use getInstructionCost to cost Shuffle. (PR #122068)

[Luke Lau via llvm-commits]

lukel97 wrote:

Hi, after this patch we're seeing a 6% improvement on RISC-V with one of the SPEC CPU 2017 benchmarks because we're able to fold away a particularly expensive shuffle: https://lnt.lukelau.me/db_default/v4/nts/profile/13/91/88

So just want to say thanks, nice find!

https://github.com/llvm/llvm-project/pull/122068