Fetch updates incrementally, but as often as allowed

[LosD]

Currently all data is re-fetched every time, and therefore updates needs to be quite seldom, both to avoid triggering rate limit and to keep the ES index small (deleted documents will linger until index is optimized).

To avoid this, the GitHub ETag header and "since" parameter can be used. To respect the GitHub API, the X-Poll-Interval response header should be heeded.