Add support for streamed chunks when making large queries

[msohcw]

Currently I've written a system that takes advantage to requests streaming capability to avoid loading all of the chunks into memory at once. Generally this seems faster and better than what is currently provided which loads all the chunks immediately. Could support be added for a generator that lazily loads chunks from buffer?

[mpas]

I have a need to export a large number of rows and seem to run in the same issue that you have solved. Could your solution be used to export these large numbers?

[msohcw]

Essentially, I wrote a wrapper around InfluxDB and extended it with a lazy_query method. I then used requests in-built streaming capability [1] to stream HTTP requests to InfluxDB. I would then return the stream as a generator.

[1] http://docs.python-requests.org/en/master/user/advanced/

[sebito91]

Fixed in #753