Benchmark using TPC-H, TPC-DS

[philippemnoel]

What
Alongside ClickBench, the most popular OLAP benchmarking framework is probably TPC-H. It's a much older benchmark, and is trusted by the older enterprises.

DataFusion already has native support for TPC-H, so I think we could add support for it really rapidly. While I think it's fine to ship without this, adding it eventually will probably really help larger enterprises think highly of ParadeDB and be confident about the maturity of our product.

There is a benchmarking framework/tool for it called HammerDB, perhaps it is useful?

Why
^

How
^

[philippemnoel]

I have started doing this, I just need to finish generating the dataset

[philippemnoel]

For these two TODOs, we'll implement them in a separate PR

• Test with various combinations of heap vs parquet tables -- EDIT: Configured, but there's a bug in datafusion-federation, which we need to fix
• Test on the 100GB generator on a c6a.4xlarge instance

[philippemnoel]

datafusion-contrib/datafusion-federation#37