Mlens

- Possible to set environmental variables
- multiprocessing default backend
- ``spawn`` as default start method for parallel jobs (w. multiprocessing)
- Possible to specify ``y`` as partition input in clustered subsumable partitioning
- Minor bug fixes
- Refactored backend for streamlined front-end feature development

Updates
- Prediction array dtype option (default=float32)
- Feature propagation
- Clustered subsemble partitioning
- No memmaps passed to estimators (only ndarray views)
- Threading as default global backend (changeable through mlens.config.BACKEND)
- Global configuration (mlens.config)
- Optional specification of temporary directory
- Scoring exception handling

Increased array checks, update of visualization APIs.

- accepts supervised transformation
- ensure training set is a view for K=2 and no preprocessing
- bug fixes model selection

Initial stable release.