Add fast download using elasticsearch's scan/scroll

[dr0i]

See #176 .

Makes use of elasticsearch scroll and scan. Fits into the API, so content negotiation is possible. Since there are some memory issues in play 2.3, e.g https://stackoverflow.com/questions/28866981/how-to-chunk-responses-in-play-without-running-out-of-memory, some workarounds are built into the code:

• only one scroll is allowed at one time.
• as less memory is available as longer pauses are enforced between getting new data.
• if free memory exceeds a threshhold (600 MB), the download is canceled

It is my experience that Promise is leaking (see hbz/lobid#188) though the memory consumption constraints mentioned above should give us enough time to react if the play instance is going to break not having enough RAM.

[acka47]

@dr0i What about this open pull request, did you forget it?