[Numpy 2.x] No longer allow conversion from numpy scalar type to pyth…

…on scalar type in consteval

This actually leads to inaccuracy during some operations, so don't try
to outsmart the developer there.

pythran/conversion.py

@@ -137,7 +137,7 @@ def to_ast(value):
     if any(value is t for t in (bool, int, float)):
         return builtin_folding(value)
     elif isinstance(value, np.generic):
-        return to_ast(value.item())
+        raise ToNotEval()
     elif isinstance(value, (numbers.Number, str, bool, type(None))):
         iinfo = np.iinfo(int)
         if isinstance(value, int) and not (iinfo.min <= value <= iinfo.max):