[llvm] Port `NVPTXTargetLowering::LowerCONCAT_VECTORS` to SelectionDAG (PR #120030)

[Matt Arsenault via llvm-commits]

arsenm wrote:

> > Given that the original code was dead code, are we still interested in merging this?
> 
> We'd still be interested in deleting the dead NVPTX code.
> 
> However, since it doesn't look like there would be any users of `SelectionDAGLegalize::ExpandConcatVectors()`, I doubt we'd want to merge since it would only amount to moving the dead code around within the codebase.


It should still merge to the generic code. AMDGPU has essentially the same code in custom lowering (and I'm sure many out of tree targets do the same)

https://github.com/llvm/llvm-project/pull/120030