Iterator

Iterator are implemented using Rayon trait UnindexedProducer, the tasks are splitted in split method. The main loop is happened in fold_with. Iterator can be used to implement elementwise or broadcast elementwise calculations. Usage can be found at here and here

Known issue

1. Current iterator like ParStrided, the fold_with method doesn't have any looping logic. Maybe we can write same logic as ParStridedZip in fold_with.

2. When outer loop size is 1, there will be no parallelism because the tasks splits based on the outer loop size.