choose_n_farthest_points number type

[mglisse]

choose_n_farthest_points currently hardcodes a number type of double. I didn't try, but I assume it would fail if we tried to run in on Epeck_d points.

Euclidean_distance manages to be generic, and I think we should do the same for choose_n_farthest_points.

[VincentRouvreau]

If I have a look to https://github.com/GUDHI/gudhi-devel/blob/7e05e915adc1be285e04eb00d3ab7ba1b797f38d/src/Subsampling/include/gudhi/choose_n_farthest_points.h#L65-L74 DistanceOutputIterator is an output iterator, which means DistanceOutputIterator::value_type is not known (cf. std::back_insert_iterator). Point_range is a range over CGAL points (Epick_d or Epeck_d), so we can have access to Point_range::value_type (will be a Point_d), but not to Point_range::value_type::FT (Point_range::value_type::FT_ is private). dist_to_L should be a std::vector< FT >, but I don't know how to get this type, maybe by using decltype on a "fake" distance computation. https://github.com/GUDHI/gudhi-devel/blob/7e05e915adc1be285e04eb00d3ab7ba1b797f38d/src/Subsampling/include/gudhi/choose_n_farthest_points.h#L93 Finally, as this comment claims, Epeck_d::FT has no infinity value, but this is quite convenient in our case to prevent the user the distance was not computed. We have to find how do so differently.

[VincentRouvreau]

A first attempt, that is not satisfying enough in this branch