Nltools

*New Functionality*
- added ISC output to align output.
- added new function to fetch pinel localizer dataset
- added smooth method to Brain_Data Class
- added find_spikes method to Brain_Data Class

*Bug Fixes*
- updated align tests

*Notes*
- dropped support for Python 2.7. Will no longer be running test suite.

*New Functionality*

- Can now perform multi-region and/or searchlight based prediction using `Brain_Data.predict_multi`
- `Brain_Data.predict` now support multi-class classification
- New interactive plotting based on `nilearn` using `Brain_Data.iplot()`
- plotting functions have been renamed to remove camel case (e.g. `plotBrain` -> `plot_brain`)
- `Brain_Data.plot` now integrates old plotting functions in method call using argument `view`. Still defaults to axial slices but can optionally display, `glass`, `mni` (for ortho slices) or `both`
- new `distance_correlation`, `procrustes_distance`, `jackknife_permutation` functions in `nltools.stats`
- `expand_mask`, `collapse_mask`, `create_sphere` can now take custom masks
- `Brain_Data.standarize` gains new axis argument

*Bug Fixes*
- Fixed issue in boolean indexing with ROC forced choice computation
- Fixed edge case bug in `fdr`
- `holm_bonf` behaves like `fdr` now and fixes incorrect pvalue bug
- `check_numpy_square_matrix` validates properly now
- fixed issue in failed arma regression if `statsmodels` was not installed
- fixed issue in failed interactive plotting if `ipywidgets` was not installed

*New Functionality*
- Added new 2-D matrix permutation to correlation_permutation()

*Bug Fixes*
- Updated MANIFEST to include resources with pypi
- fixed compatibility issues with python 2.7 and joblib. Using older version for now.
- refactored tests to be cleaner with fixtures.

*New Functionality*
- New regression method for Adjacency objects (univariate and Adjacency decomposition)
- faster t-tests (permutation and standard)
- within cluster mean method for Adjacency objects
- similarity to distance conversion for Adacency objects
- new function to convert ROI values back into brain_data instances

*New Functionality*
- Brain_Data now has scale method to be consistent with spm and fsl

*Bug Fixes*
- Big refactor to design_matrix class.
- Updated documentation.

*New Functionality*
- added labels to Adjacency Class
- refactored regression module (includes robust and AR models)
- updated threshold methds
- refactored design_matrix class. Lots of new functionality
- refactored trim and winsorize into single function
- permutation tests now default to two tailed

*Bug Fixes*
- plot_mds
- upsample
- importing lists of 4d data.
- transform_pairwise
- Design_matrix many bug fixes
- conflict with matplotlib and nilearn versions