Quick clickbench with a smaller dataset

### Is your feature request related to a problem or challenge?

For measuring the performance improvement of #11827 , some extended queries with `more complex udaf(like median, approx_median)` + `high cardinality group by` are needed #12438 .

But I found, such queries can't run successfully to get the result in my local. After debugging, I found it is due to their large intermdiate results which will full memory rapidly, leading to swap or oom...

However, when I run it in a subset with only 15% of the whole clickbench dataset, they can finish successfully and reflect the improvement https://github.com/apache/datafusion/pull/11827#issuecomment-2347092811

I think maybe we need a clickbench with the smaller dataset (like tpch 1, tpch 10...) in some situations.

### Describe the solution you'd like

Support to generate a samller dataset of the whole clickbench dataset, and we can run queries on it.

### Describe alternatives you've considered

_No response_

### Additional context

_No response_

Start by reviewing the existing ClickBench dataset generation and the performance context in #11827 and #12438. Define how a smaller dataset should be produced, then verify that the complex UDAF and high-cardinality group-by queries can run on it without exhausting memory.

Written by the indexing model from the issue text.