[llvm] [IR] Implementation and lowering of bitinsert and bitextract (PR #200605)

[Alexis Engelke via llvm-commits]

https://github.com/aengelke commented:

Looks ok, but the codegen for variable bitinsert is still abysmal and I still think this should not use rotates (which I believe are the source of the bad codegen). I would have also preferred codegen changes to land separately from IR changes, but whatever...

https://github.com/llvm/llvm-project/pull/200605