use dask-cudf to utilize multiple GPUs

[jangorecki]

cudf uses only single GPU, thus it would be useful to employ dask-cudf rather than just cudf.
https://blog.dask.org/2019/01/29/cudf-joins

[jangorecki]

this can unfortunately cause other issues, for example rapidsai/cudf#3363 (comment)

[jangorecki]

As pointed out by @datametrician we should be also able to use off-vmemory data storage with dask-cudf, then it even make sense to use dask-cudf for a single GPU.

[jangorecki]

waiting for rapidsai/cudf#2288

[jangorecki]

Dask seems to be a not mandatory for spilling to main memory. Due to poorly documented setup of dask-cudf this part will be solved separately: #129

[datametrician]

Without Dask, you still only use 1 GPU instead of both of them.

[jangorecki]

@datametrician yes, I am aware of it, so the plan is to move to dask-cudf, so this issue stays open.

[jangorecki]

using dask-cudf will additionally allows to attempt 1e9 data size by using spil to disk memory feature, as explained in rapidsai/cudf#3740 (comment)

[jangorecki]

rapidsai/cudf#2288 has been finally resolved and it looks we can proceed to using dask_cudf to utilize both GPUs