Adding Dask-SQL Benchmarks

[quasiben]

As an experiment, I built a notebook with SQL versions of the queries and compared outputs to check validity:
https://github.com/quasiben/h2o-benchmarks/blob/sql-version/dask-sql-h2o-benchmarks.ipynb

The sql is based on DuckDBs solutions:
https://github.com/h2oai/db-benchmark/blob/master/duckdb/groupby-duckdb.R

(Tested with 5GB datasets)

Would you be interested in adding Dask-SQL equivalents to this repo to be benchmarked as well ? If so, we should probably confirm runs at 50GBs as well. If not, no worries -- mostly I'm happy that Dask-SQL "just worked" for the queries thus far.

cc @charlesbluca @randerzander

[ncclementi]

In case you folks need it, we do have the datasets in parquet and multiple csv files up to the 50GB case publicly available. There is info in the readme about it: https://github.com/coiled/h2o-benchmarks#public-data-on-s3