[llvm] [RISCV] Adopt SpacemitX60's scheduling model for `-mtune=generic` (PR #167008)

[Min-Yih Hsu via llvm-commits]

mshockwave wrote:

We have benchmarked SPEC2006INT of this change on SiFive X280 and observed slight regression (~1%) on the geomean. Though it's worth mentioning that we observed a larger regression on 464.h264ref, which is relatively vector-heavy. This is kind of expected, as it is well known that X280 has quite a different performance characteristics of its vector unit compared to that of SpacemitX60.

I'll see if I could collect SPEC2017 data, but in any case we don't think the regression on X280 should block this patch moving forward. Therefore, all things considered, I think this patch is in a good shape for another round of review (once I resolved the outstanding merge conflicts, that is)

https://github.com/llvm/llvm-project/pull/167008