arrow

ARROW-10551: [Rust] Fix unreproducible benches by seeding random number generator

This PR fixes all unreproducible benchmarks and tests.
1. how is the current code broken?
Current, code is reseeding after 32 MB of data, and every thread has different randomness. So no data is same as another data and totally different in different use cases.
2. how is this pr solving it?
By introducing a code that have fixed seed for all threads at the all times. and all threads are at the same random number at the given time of a single benchmark.
3. why do we need to take another dependency (i.e. why doesn't random work)?
random usees thread_rng because of the reason mentioned in 1 it is not working and thread_rng also xors with the seed given from the threaded itself. So it is mostly random at any given time, but not consistent across concurrent behavior like benchmarks and things like that. Dependency is bastion runtime's utility crate. So runtime is using it already.

Seeing that we are not using rand there a lot, we can actually remove the rand with this PR too.

Closes #8635 from vertexclique/ARROW-10551-fix-unreproducible-benches

Authored-by: Mahmut Bulut <vertexclique@gmail.com>
Signed-off-by: alamb <andrew@nerdnetworks.org>