Spatial skill per quantile

[daniel-caichac-DHI]

Hi @jsmariegaard and @Hendrik1987,

I was wondering if it would be possible when running the spatial_skill comparer to specify, besides the spatial bins, a range of quantiles.
Right now I understand that spatial_skill (given some bins) is binning the data and getting the metrics (rmse, bias, etc) for each bin, between Model and Measurements, for all the overlapping data, but what if the user only wants to compare Model v/s measurements for data above the 95th percentile (or 99th, or Nth percentile), to see if a given model performs better for extreme conditions.

The most complex part that I see right now is that the given percentile would change for each bin, as it will be variable in space.
Any thoughts on this?