Faster OpenAQ (using openaq-fetches)

[zmoon]

Some initial timings suggest we can gain speed by using s3 URL scheme instead of current HTTPS:

In [8]: %timeit -n 10 pd.read_json("https://openaq-fetches.s3.amazonaws.com/realtime/2013-11-26/2013-11-26.ndjson",
   ...:  lines=True)
311 ms ± 37.1 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)

In [9]: %timeit -n 10 pd.read_json("s3://openaq-fetches/realtime/2013-11-26/2013-11-26.ndjson", lines=True)
159 ms ± 13.5 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)

similar to #113

Originally posted by @zmoon in #116 (comment)

[zmoon]

Also was experimenting at some point and found some ways to speed up the JSON parsing, datetime in particular.