Refactor Scholar.Neighbors

[krstopro]

As argued in #254, we need to update Scholar.Neighbors in the following way:

• [x] Add BruteKNN module that implements brute-force k-NN search. Its predict function has to return {neighbor_indices, neighbor_distances}. Make batch_size an option (as in #253).
• [ ] Implement KNNClassifier.
• [ ] Implement KNNRegressor.
• [ ] Deprecate Scholar.Neighbors.KNearestNeighbors.

KNNClassifier and KNNRegressor have to take k-NN algorithm and metric options as atoms, but also allow for custom k-NN algorithms and metrics to be passed as modules and functions, respectively.