Create standardized and optimized tool for validation dataset creation

[mattgara]

A validation dataset can be optionally used when benchmarking in velox-testing to ensure results are correct, not just performant.

Currently, there is no optimized method to create this dataset, and so an ad-hoc solution typically must be used for larger scales of data.

Let's ensure we have a shared, standardized, and (most importantly) optimized method to create such results.

[mattgara]

@misiugodfrey @paul-aiyedun