[Feature Request] Enable cloud-based Model Optimization

[larsll]

To improve the speed of loading models on the car it would be an idea to perform the Tensorflow -> OpenVINO model optimization in a lambda function, rather than on the car. This should save 15-20 seconds when loading a model on the car.

Initial concept (Docker file + conversion code) is here: #9

Design decisions needed:

• Where in the upload-pipeline should it happen? Two options
  • Instead of clamscan_post; trigger on virus scan complete, then optimize and upload to target location
  • After clamscan_post; when the model is Available (maybe trigger on the change in DynamoDB?)
• Is another status in the DynamoDB desired (Available -> Optimized?)
• If processing after clamscan_post: shall initial file be kept and new file in a new location, or shall it be edited in place?

[davidfsmith]

• Optmisation should happen after clamscan_post
• Yes, another status of Optimised
• Keep both files and then we can decide which version of the files to load to the car (depending on the software on the target car)

[davidfsmith]

This change has now been merged in as part of PR #9