In the k-NN functions it should be possible to consider other forecast horizons (besides t+1)

[javiag]

Right now, the functions only forecast for t+1, but the functions should include a parameter for specifying other forecast horizons. The name of the parameter can be f_horizon. If, for example, f_horizon is 2, the k-NN will forecast t+2 (and not for t+1).

All the k-NNs should be adapted to include this parameter.