Object-storage support

[vinjana]

Read input from S3 and write output to S3

Issues:

• Roddy checks input files for existence, maybe in checkExecutibility() from workflows. Provide a way to transparently check input, no matter whether file or S3 object.
• Three alternative access modes: (1) URI (file://..., s3://..., others are in principle possible like hdfs://...) and let the job handle it, (2) pre-download the object to a temporary local file before the actual jobs starts, (3) provide a stream to download the complete file (similar to decompression but not refering to file:// but s3://; decompression may additionally be necessary).
• Automatically upload declared output files (those that are job-outputs) to bucket.

Related existing concepts:

• File: May also be an S3 object under the hood (generalize to DataResource or so)
• FileSystemAccessProvider: Could also be S3 (generalize to DataResourceAccessProvider or so)

TODO:

• Collect more issues, like those above
• Identify composition into minimally viable products (MVPs)
• Create issues to create MVPs