Nltools

New Functionality
- added divide dunder method to Brain_Data and Adjaceny
- added right dunder methods to Adjacency
- added axis argument to Brain_Data.std()

New Functionality
- new interactive plot to view components from decomposition analysis
- new median method
- can now extract median and PCA components from a masked Brain_Data instance
- can now perform KernelPCA and DictionaryLearning with Decompose method.

Changed API
- many `method` keywords have been changed to `metric` for consistency (e.g., distance methods)
- extract_roi Brain_Data method now has option to resample target mask to source data

Bug Fixes
- fixed the double plot in jupyter notebooks
- removed nans from regression standard error
- fixed issues with labels in several Adjacency plots (mds, silhouette)
- fixed issue with float and integer types in all dunder methods in Brain_Data & Adjacency
- fixed roi_to_brain function, which can now accommodate most input data types
- changed how empty labels are stored in Adjacency
- added mask argument to check_brain_data
- refactored and updated apply_mask, new flag to resample mask to

Problems uploading to pypi - see version 0.3.15 for full release notes.

**New Functionality**
- added plot saving functionality for full, glass, and min views
- predict now uses balanced_accuracy as default for classification accuracy

**Bug Fixes**
- fixed bug with glover_hrf function which will impact convolution with design_matrix
- fixed numerous bugs with adjacency class
- fixed multiple mugs in permutation tests
- fixed bug with 3mm brain mask
- fixed bug checking for isometric voxels which was causing plotting errors
- fixed bug in plotting save
- updated sklearn requirements which was causing bug with backwards compatibility for balanced_accuracy
- fixed bugs with hdf5 with adjacency instances

New Functionality

- Added ability to read and write Brain_Data objects to hdf5. This is WAY faster than nifti and is the recommended method for large datasets.
- Randomise - can now run fast permutation testing similar to FSL.
- Add options for holm-bonf to ttest and randomise
- Added ability to estimate Social Relations Model using Adjacency Class.
- Added new SimulateGrid class for generating random data for simulations in 2D

*Bug Fixes*
- fixed fetch_localizer
- fixed bug in Brain_Data.smooth where original class instance attributes weren't being propagated.
- cleaned up code for distance_correlations
- fixed bug with Adjacency when NaNs were present on diagonal.
- fixed travis bug.
- fixed plotting bug

*New Functionality*
- Can now multiply vectors to brain data instances. useful for making contrasts

*Bug Fixes
- Updated documentation
- Fixed readthedocs conda environment