[llvm] [RISCV] Expand constant multiplication for targets without M extension (PR #137195)

[Philip Reames via llvm-commits]

https://github.com/preames commented:

Can you expand your review description with the motivation, and a bit of justification?  In particular:
* Why is unconditional expansion the right answer?
* Why not use the simpler sum of powers of two?  How much does the complexity of the factorization and NAF lowering really buy us?  Feel free to cite a paper/blog if this is "well known".

https://github.com/llvm/llvm-project/pull/137195