Allow partial download of imagesets

[AbhiDhir]

Overview

• Allows the option for partial download of imagesets rather than requiring full download \
• download_full_imagesets is a new optional parameter in config
  • if set to true, behaves how it used to (all imagesets are fully downloaded into cache)
  • if set to false or not present, setSizeFilter must be set to values indicating how many images from each imageset are required
    • This is currently up to the plugin to check for and can be seen in the plugins repo

Notes

• The current implementation for asynchronous download may be done incorrectly or not as efficiently as possible,
  • I will work more on this by doing time testing
• This pull request goes along with the request in the ravenml-dataset-plugins repo

[jacobDeutsch10]

Just tested by making a small dataset and it worked for the most part. The only issue i saw was that it if an image-id doesn't have an associated file for a given prefix/suffix pair it will download an empty file for that image-id and prefix/suffix pair.

in the tfrecord plugin the associated_files are:
associated_files = [ ('meta_', '.json'), ('image_', '.png'), ('image_', '.jpg'), ('image_', '.jpeg'), ('bboxLabels_', '.csv') ]
but since the imageset i used only contains files with for ('meta_', '.json') and ('image_', '.png'). we get empty files for each of the remaining pairs.

I think this could cause some issues if this list was ordered differently