TPC-DS at different scale factors

My understanding from #25 is that the data generation itself is somewhat random so it should be static rather than generated on the fly, so would a PR adding a fixed version of SF=10 be welcome?

My main motivation is diving into [this](https://github.com/apache/datafusion/issues/17494) DataFusion issue I've filed a while ago, right now my main way to reproduce the issue is through another benchmarking setup but that is a bit convoluted and it'll be really nice to use the built-in setup in the main repo.