Shard spaces (rebased variant of PR #23)

[Civil]

Sorry, I've lost the repo where it was and I don't know how to rebase PR from deleted repo, so I've decided to create new one.

[Civil]

Please note, that I don't have InfluxDB setup right now, so I was unable to test the code. The only tests I've performed - run pylint and check that there are no syntax errors.

[Dieterbe]

@liyichao you were a big fan of the previous PR, could you test this? i'm super busy right now :(

[liyichao]

I setup a test environment, where I use another docker image with Civil's pull request, but it reports 500 when I direct grafana to use that as a backend.

graphite-api's log seems useless, so I try to run app.py manually using Flask's app.run(debug=True),
but failed to do so because #41

Also, influxdb 0.9's shard spaces will be changed, which I'd like to wait for before setuping retention policies.

Currently, influxdb can not have multiple graphite-servers making it difficult to set up a totally separate testing environment for me. Because we use a forever shard space without any down-sampling and retention policies, what I can test is also limited.

[liyichao]

@Civil I use graphite-api and here is my schema:
schema:
- ['', [['', 10, '30d'],['10_mins_rollup.', 600, '700d']]]