Merge AggregatedSkill and SpatialSkill classes?

[jsmariegaard]

Would it be possible to have a single Skill class having a xr.Dataset to cover both the retun of skill(), mean_skill() and spatial_skill() ? We could then get rid of the MultiIndex 🤔. The ds should contain the meta data from the Comparer class which could be used for filtering etc (e.g. DA-station vs validation station). The coords could be observation, model, x, y, z, etc (depending on the by argument). The dimensions would also depend on the by argument and in case of the gridded skill also be x and y. I guess the AggregatedSkill is just a special case of GriddedSkill with only a single "point" per model/obs...

Thoughts? @ecomodeller ?