1.0.0

[freeman-lab]

This PR tracks what will be a 1.0.0 release. It reflects a major refactoring, including many breaking API changes, thus the 1.0.0 label! Below are some of the primary changes (nearly all are complete):

• support for both local analysis (w/ numpy) and distributed analysis (w/ Spark) via a common API
• a design that can support a variety of computational backends
• py.test for unit tests
• use S3 for example data so it's not included in the repo
• no more executables or standalone scripts to simplify deployment and installation
• more consistent and python-friendly style conventions
• updated documentation
• python 3 compatibility

Another major change is breaking out functionality into a more modular collection of smaller libraries that can be independently versioned, updated, and tested. In this reorganization, thunder will be responsible only for the core operations of loading and manipulating image and time series, but can be composed and used alongside with several additional libraries, which themselves can be used for either local or distributed purposes.

Analytics

• thunder-extraction - feature extraction from image sequence data
• thunder-registration - registration for image sequence data
• thunder-factorization - matrix factorization methods
• thunder-regression - mass-univariate regression modeling

Utilities

• bolt - ndarray-like interface for spark
• station - context manager for computational engines
• checkist - argument checking
• showit - simple display of images and tiled images

[freeman-lab]

finally merging this one in! thanks for the hard work everyone!