Complete initial preprocessing pipeliine

[thenofreeman]

Should:

• extract data from given data set into parseable json (done)
• take the extracted data as give as input to VLM with cropped image for prediction
• receive prediction for location, store it in the appropriate key, and be able to compare it to the true classification
• all this data should be available to be piped into a database which we can query with pynamo.

don't store it in the database (unless for testing) until we are reasonably confident that the format is working and classification predictions are sound enough.