Wenlei Xie

cc @dracifer

thanks @myzha0 > Seems like a easy hack would be to just make a wrapper to ListColumnCpu that wraps _from_pysequence [here](https://github.com/pytorch/torcharrow/blob/9aecec2501f565f08a2556863386f94cccc8ebed/torcharrow/velox_rt/list_column_cpu.py#L61). Yeah that requires two copies (arrow -> Python objects...

@pearu, @ezyang > I imagine that the undefined fill-value feature would be also useful in manipulating sparse tensors representing weighted adjacency matrices from the graph theory or in other applications...