Implement a benchmark for extracting arrow statistics from parquet #10606

Is your feature request related to a problem or challenge?

Part of #10453

Part of #10453 is to "efficiently" extract statistics from parquet files. The current code is likely quite bad, but we have no benchmarks that actually show that or allow for us to get better

Describe the solution you'd like

I would like to create a benchmark, run like cargo bench ... that does:

1. Creates a parquet file that has 100 row groups once
2. Read the parquet metadata out
3. in the loop extracts the statistics from the metadata into Arrow statistics (using the API introduced in feat: API for collecting statistics/index for metadata of a parquet file + tests #10537) -- not it shouldn't including reading the metadata from the parquet file

I suggest four benchmarks with different types:

1. UInt64
2. f64
3. String
4. Dictionary(Int32, String)

Describe alternatives you've considered

No response

Additional context

No response