[FEA] BSDG: Add batch option

BSDG (Billion Scale Data Generator)
Open-sourced version in cuvs bench currently flushes the full dataset to disk on generation.
For very large datasets, users may want to generate in partitions.

If user gives a batch_size and a batch_id we should just return/save that specific batch.

# returns rows between batch_size * batch_id ~ batch_size * (batch_id+1)
batch = gen_batch(total_rows, batch_size, batch_id)

This is straightforward because we know the number of points per cluster, so we should be only able to generate the necessary clusters that have vectors belong to that partition of the full dataset.