Address dask_cudf.read_csv chunksize deprecation

[mroeschke]

xref #4271

chunksize was deprecated in favor of blocksize

Also removed an unsupported chunksize in a cudf.read_csv call

[acostadon]

Looks good

[rlratzel]

Thank you very much, @mroeschke !

So is blocksize simply a rename of the old chunksize arg, or are there bigger differences? I'm just asking in case we also need to update get_chunksize() to return something that's computed differently.

[mroeschke]

So is blocksize simply a rename of the old chunksize arg, or are there bigger differences?

I believe the change is just a pure renamed based on rapidsai/cudf#12394, but tagging @rjzamora just to confirm

[rjzamora]

I believe the change is just a pure renamed based on rapidsai/cudf#12394, but tagging @rjzamora just to confirm

Yeah, I believe someone arbitrarily chose the name chunksize in the early days of dask-cudf, even though the upstream dask.dataframe function was already using the name blocksize for the exact same purpose. When we started closing the gap between the dask_cudf and dask.dataframe APIs a couple years ago, we deprecated chunksize in favor of blocksize.

[rlratzel]

Thank you!

[rlratzel]

/merge