Simple parallel processing implementation

[d-chambers]

This post highlights an idea to better support simple parallel processing in DASCore. Ideally, we would implement a map_to method on the BaseSpool. It works like this:

import dascore as dc 

spool1 = dc.spool("input_data_source")
spool2 = dc.spool("output_data_location")


def patch_proc(patch):
    """function to perform patch processing"""
    ...
    return out

spool_out = spool1.map_to(patch_proc, spool2)

It would support passing in a client which has a map method (to support parallelization with dask, ray, threading, processing, etc.) as well as a chunksize argument.

from concurrent.futures import ProcessPoolExecutor

# create a process pool
client = ProcessPoolExecutor()

# each process gets a spool with 10 patches to iterate through. 
spool_out = spool1.map_to(patch_proc, spool2, client=client, chunksize=10)

This works because spools are lazy and can be serialized with pickle.

To make this work though, we need a few more spool methods:

1. spool.put - puts a patch into the spool. Not all spools have to implement this
2. spool.split - split the spool into sub-spools with certain number of patches in each (slicing already does this)

Thoughts? Different ideas?