Multi processing implementation for Deep Learning PODs

[chrise96]

Tested and works tot postprocessing POD

• Multi processing implementation works with DAG cvt-pipeline-small_multiprocessing
• Part of the batch processing implementation is in the comments of retrieve_images.py

TODO for later

• Postprocessing doesn't work, need to make implementation for glob on coco_instancesresults*.json files
• Fix when you want to restart DAG. Idea by Thomas:

file_names = read_csv(path_to_csv)
try:
    for file_name in file_names:
        process(file_name)
except Exception as e:
    # Clean up processed files to prepare for restart
    for file_name in file_names:
        # Do something to get the path to the processed file
        processed_file_name = filename += "processed"
        os.rm(processed_file_name)
    # Re-raise error to log trace back, and bring task to failed state 
    raise e