[lld] [llvm] [RISCV] Teach RISCVMergeBaseOffset to merge %lo into load/store folding arithmetic (PR #185353)

[via llvm-commits]

LiqinWeng wrote:

> Yes, it does seem reasonable to look at (and delete) the `addi` coming into any register-relative add, to fold it into register indexing. I don't think this specific sequence needs to be run past the ABI group, as it's basically a generalisation of (1).

Pls see issue: https://github.com/llvm/llvm-project/issues/185586

https://github.com/llvm/llvm-project/pull/185353