[llvm] [RISCV] Strength reduce mul by 2^M - 3/5/9 (PR #88993)

[Philip Reames via llvm-commits]

preames wrote:

> There's a bug here I need to fix, will post a revised version later this morning.

Actually, it appears the original code was correct, and I actually had a redundant check, but skeptical review here is appreciated.  Overflow logic makes my head hurt, so a skeptical second look at this is appreciated.  


https://github.com/llvm/llvm-project/pull/88993