Wsknn

- updated session selection algorithm, added `weighted_events` parameter that allows to weight the closest neighbours by the vector of weights,
- new tutorial with comparison of a different recommendation settings,
- updated documentation.

- updated scoring selection,
- package is stable.

- added `return_events_from_session` parameter to force algorithm to return items the same as in a given session if there are no other neighbors,
- `recommend()` method resets settings if provided (instead of `fit()`).

- `predict` method changed to `recommend`, accepts only List of Lists as the input,
- refactored scoring,
- added `recommend_any` parameter to force algorithm to return k-recommendations. If there is less than k recommendations, and this parameter is set to `True`, then algorithm returns missing products from the product pool at random,
- added `.dev` suffix to package version, to pinpoint the fact that package is still unstable.

- Changed API with metrics.
- Fit method takes hyperparameters.
- Changed API for precision calculation.
- Session slicing bug corrected.

- Make predictions from the List of user actions,
- Include required event in the `predict` method parameters,
- Let user decide if scores are calculated from a sliding window or fixed-length size window (the latter protects from the counting the same results multiple times).