Record index of job inputs and outputs in S3

[Foxcapades]

For debugging/transparency, we should write an index file to S3 for a given job's inputs and outputs.

For ease of implementation, this doesn't need to be a singular file, we could have something like inputs.json and outputs.json which contain arrays of file names.