[llvm] [RISCV][CostModel] Add getCFInstrCost RISC-V implementation (PR #65599)

[Philip Reames via llvm-commits]

preames wrote:

> LGTM, but please wait for sign off from @preames since he was the one that asked about performance impact and hasn't had an opportunity to weigh in yet.

I have no objection to this landing.  I'm not completely convinced by the data shared so far - in particular, no performance or dynamic instruction statistics were shared - but I'm fine with this landing and we can revert if perf data shows the change in vectorization isn't actually positive.  

https://github.com/llvm/llvm-project/pull/65599