ignore most distant curves - outlier handling

[martin-raden]

As a crude and most basic outlier handling, one can ignore the k curves that have on average the highest distance to all other curves. (most simple)

A bit more problem dependent: set a cutoff instead of a fixed number. Here: use % of pairwise distance range to enable a [0,1] scaled value selection.