[FEA] Benchmark Jaccard with alternative distributions

**Is your feature request related to a problem? Please describe.**
https://github.com/rapidsai/cudf/pull/13669 adds the Jaccard index feature. It also includes benchmarks using normally distributed data. However, realistic data is likely to be more geometric in nature and have longer-tailed distributions with a small number of much larger strings.

**Describe the solution you'd like**
We should add a benchmark with a geometric (or other skewed) string length distribution.