Merge S3 Commuter Into Commuter, following bookstore organization

[rgbkrk]

We now have a few discrete prefixes associated with one common bucket, granted via bookstore:

• /workspace
  • bookstore is auto-archiving notebooks for users into this prefix
• /published
  • Not yet in use (pending nteract UI & bookstore config work)
• /papermill
  • Created by @mseal et. al. for job artifacts, not really part of bookstore; just the pattern being used

We currently run two commuters. One for S3 (coded to work against one bucket and prefix at a time) and one for EFS. Given that we're going to take commuter off of EFS we need to have a transitiion over to the S3 backed version (what's now going to /workspace). Here's what I propose (and what @stormpython, @mpacer talked about yesterday):

• create an /api/s3-contents/${BUCKET}/${PREFIX}/${PATH} endpoint that only allows particular safe listed buckets
  • We'll also have to make /s3-files
• Create new pages targeting each of the above buckets with names:
  • pages/job-artifacts.js - /papermill/scheduler-output
  • pages/published.js - /published
  • pages/workspace.js - /workspace
• Create a way to configure which bucket & prefix goes to each of those, likely sourced from new COMMUTER environment variables

[mpacer]

In order to make the frontend pages know what prefix will actually be used, we will probably want to expose configuration at a higher level in the package.

Right now, there's an implicit coordination happening via the COMMUTER_PORT env variable, which has been the only place where configuration would matter before. That's why this hasn't been a pressing need up to this point.