Stats fixes and enhancements

[ejolly]

These are bug-fixes but many are enhancements that would be likely candidates for moving into burnt-ends. We had also discussed creating a stats output class #172 which should be straightforward using the new Python dataclass decorator.

Relevant issues:

• Clean up predict algorithms #182
• Refactor ROC module #183
• Stats output class #172
• add permutation tests to predict and roc_plot #176
• add balanced option to predict classification #177
• make lassopcr predict option more clear in documentation #148
• add 2d_permutation for Adjacency ttest methods.  #244
• stats.threshold usage is confusing #304
• How to handle interaction between multiple comparisons and 1-tailed tests #315
• Add FastSRM Algorithm #345
• add interpolation option to append #354
• Adjust sigma estimation for standard error computation in regress #287
• Replace custom stats.procrustes with scipy import #417
• Add different similarity metrics to stats.isc() function #395
• align with SRM model doesn't allow different numbers of samples even if num features is the same #410
• stats.upsample should use .items() not .iteritems()  #418