Use an explicit skip list for columns ignored during auto-grouping

Instead of deriving the ignored set from the complement of
SCALED_AESTHETICS, list the non-aesthetic/special columns directlyl.
This allows custom geoms and stats to define aesthetics that affect
the grouping.

Author: has2k1

plotnine/layer.py

@@ -577,7 +577,12 @@ def add_group(data: pd.DataFrame) -> pd.DataFrame:
         return data
 
     if "group" not in data:
-        ignore = data.columns.difference(list(SCALED_AESTHETICS))
+        # Exclude some "special" aesthetics from contributing towards
+        # the grouping.
+        # label - geom_text, geom_label
+        # geometry - geom_map
+        # where - geom_ribbon
+        ignore = ["PANEL", "label", "geometry", "where"]
         disc = discrete_columns(data, ignore=ignore)
         if disc:
             data["group"] = ninteraction(data[disc], drop=True)