Build test suite of L8 data about rio-cloudmask

[perrygeo]

One of the challenges to this task is that the cloudmasking algorithm must operate per-scene as several of the operations require global percentiles. So we can't do clipped subsets without biasing the results.

The other option is to do reduced resolution overviews but there could be some bias from resampling and scale issues.

Final option is to write a script to be run on an EC2 instance that pulls full datasets from s3://landsat-pds, runs the operational test suite. The results could be downsampled/clipped and transferred locally for inspection.