[Python] Implement Scanner::from_fragments() to initialize a scanner from sub-set of fragments

[eddyxu]

Problem Statement

It will be useful for fine-tuned distributed scan. a master node can decide the plan that assigns fragments to each distributed node. Each distributed worker could scan the fraction of the dataset accordingly.

[GallagherCommaJack]

ideally the API for this accepts an iterator of fragments (vs a list)

[changhiskhan]

This means adding a fragment_ids setter or smth to Scanner and modify the Scan IO node to take this list as a parameter ?

[eddyxu]

Yes, this could be just a constructor from Scanner, takes in the fragments, and build a scanner in one worker.

Can follow the pattern like https://arrow.apache.org/docs/python/generated/pyarrow.dataset.Scanner.html#pyarrow.dataset.Scanner.from_fragment