vote: migrate mrpowers-benchmarks to mrpowers-io

[MrPowers]

mrpowers-benchmarks are starting to get referenced and used, see here for an example.

We could start publishing some better results for the data community here: https://www.mrpowers.io/benchmarks

Here are/will be the key advantages of the mrpowers-benchmarks compared to the h2o/DuckDB benchmarks.

• use Parquet files instead of CSV
• compute total query runtime, not just the query runtime when data is loaded into memory
• present the results in a more representative way (if you show the cumulative runtime for 10 queries, then a really slow query can greatly bias the overall result)
• add new queries including compound queries like filter, group by, and then join
• add synthetic datasets with complex column types like StructType, MapType, and JSON strings