Add capability for downsampling high resolution imagery, or facial recognition and blurring algorithm #382

See responsible-usage-guides/drone#3

• This is pretty critical as high resolution imagery becomes an reality using cheap hardware.
• Many may not realise that publishing this imagery publicly could be a significant risk for personal identification / privacy.
• The new OAM uploader is a multi-stage process, with each stage completed via a container in an Argo workflow.
• It's definitely in the real of possibility to add additional steps for either (1) downsampling (2) face blurring for to the upload workflow.
• Panoramax already has some facial recognition and blurring algorithm in place to handle street level imagery, so perhaps we could work with that.
• The best case is default face blurring. A worse case is simple downsampling of imagery that is higher resolution than 3 cm/px GSD, as this is the point at which faces start to become recognisable in imagery.

Additional info

• Panoramax algorithms: https://gitlab.com/panoramax/server/sgblur
• Perhaps fAIr could be involved here, but that adds significant additional complexity and coupling of tools that I'm not a fan of.