[llvm] [DAGCombine] Remove oneuse restrictions for RISCV in folding (shl (add_nsw x, c1)), c2) and folding (shl(sext(add x, c1)), c2) in some scenarios (PR #101294)

[via llvm-commits]

LiqinWeng wrote:

> For https://godbolt.org/z/o8GdMKrae, the gcc code for `add_shl_moreoneuse_store` looks like a direct translation of the original code. Is something in the middle end changing the IR and making it produce worse code?

I have tried doing it in the instcombine phase before, and there is a trend of performance degradation for some test cases on some architectures~Could you help me take a look at this issue：https://github.com/llvm/llvm-project/issues/112830

https://github.com/llvm/llvm-project/pull/101294