[llvm] [RISCV] Always expand zero strided vp.strided.load (PR #98901)

[Luke Lau via llvm-commits]

lukel97 wrote:

> It's unfortunately we loose the smaller EVL here, but this looks like the right pragmatic tradeoff at the moment.

My thinking is that we only lose the smaller EVL when we have optimised zero stride loads, so a larger EVL shouldn't matter anyway. 

https://github.com/llvm/llvm-project/pull/98901