does it make sense to add concurrent benches?

**Which part is this question about**
arrow-rs or parquet

**Describe your question**

Right now the parquet/arrow benches are mostly single-reader / single-query shape. Wondering if it make sense to add concurrent read benches to answer questions like how multiple readers interact on the same file.

Start by reviewing the existing parquet/arrow benchmarks and how they model single-reader and single-query workloads. Determine what concurrent readers sharing one file should measure and how completion would demonstrate useful interaction data without changing the benchmark scope.

Written by the indexing model from the issue text.