[llvm] [RISCV][TTI] Reduce cost of a build_vector pattern (PR #108419)

[Luke Lau via llvm-commits]

lukel97 wrote:

Just a quick update on the benchmarking, with LTO the regression disappears. 511.povray_r is actually significantly faster with LTO, by about 5.5%. I'm rerunning the benchmarks again with LTO, but not with -Ofast. -ffast-math seems to cause the tests to fail

https://github.com/llvm/llvm-project/pull/108419