[EPIC] Add other benchmarking datasets

Right now we can run benchmarks on three different datasets:

• TPCH
• TPC-DS
• ClickBench

However, it would be cool to add some more, for example:

• Variants of existing benchmark, like sorted ClickBench or sorted TPCH.
• Other datasets, like h2o or imdb.

We have good foundations here for holding more datasets easily, the process typically is:

1. Create the Rust code that prepares the dataset in https://github.com/datafusion-contrib/datafusion-distributed/tree/main/benchmarks/src under a prepare_<dataset>.rs name.
2. Create the .sh script that calls the Rust code with the appropriate config and generates the dataset in https://github.com/datafusion-contrib/datafusion-distributed/tree/main/benchmarks under a gen-<dataset>.sh name, under different scale factors or configs:
   • If the dataset allows scale factors, the dataset should be generated in testdata//sf/...all dirs with parquet tables...
   • If not, we can just do testdata//...all dirs with parquet tables...
3. Place the SQL queries used for benchmarks in https://github.com/datafusion-contrib/datafusion-distributed/tree/main/testdata under <dataset>/queries/q<number>.sql
4. Wire up the new dataset in https://github.com/datafusion-contrib/datafusion-distributed/blob/main/benchmarks/src/run.rs (queries_for_dataset)

The result is that running a command like this:

WORKERS=8 ./benchmarks/run.sh --threads 2 --dataset <dataset>[/<sf100>]

Should end up benchmarking the new dataset

The task is big and will need to be broken down in several sub issues that people can contribute to independently:

• https://github.com/datafusion-contrib/datafusion-distributed/issues/631
• https://github.com/datafusion-contrib/datafusion-distributed/issues/633
• https://github.com/datafusion-contrib/datafusion-distributed/issues/632

Start by reviewing the existing dataset preparation files under benchmarks/src, generation scripts under benchmarks, query files under testdata, and queries_for_dataset in benchmarks/src/run.rs. The linked subissues define the independent contributions. Done means a selected dataset is generated in the expected Parquet layout, its queries are present, and benchmarks/run.sh accepts the dataset configuration.