Overview

This PR uploads tifs that fail to process to a debug data bucket so we can figure out what went wrong even if
people sensibly follow our advice about deleting their failed projects.

Checklist

• Description of PR is in an appropriate section of the changelog and grouped with similar changes if possible

Testing Instructions

• raise an exception (any exception) at the beginning of the GeoTiffS3SceneFactory
• rebuild the batch container -- docker-compose build batch
• if you haven't in a while, rebuild the batch jar from sbt: batch/assembly
• fire up groundwork and your RF servers and create an upload with the network tab open
• copy your upload's id
• try to process your upload, pretending that this is the third job attempt: AWS_BATCH_JOB_ATTEMPT=3 ./scripts/console batch 'rf process-upload d64e5be4-bf8c-4c82-90bd-d7ea9ad4082d'
• you should see a message beginning Skipping debug data upload
• set a debug data bucket that you can write to (any bucket will do but I have a suggestion)
• run the upload processing command again
• You should see a message ending It will be available in the debug bucket for 14 days.
• check the bucket / key that that message reported -- there should be a tif there
• confirm that the environment variable for the debug data bucket matches what's in azavea/raster-foundry-deployment#255

Closes #1049