Zero-dimensional arrays should be accepted as scalars

Currently, zero-dimensional arrays get rejected on `NumericalTrait(ndim=0)`, as the latter requires the value to be `np.isscalar`. Unfortunately `np.isscalar(np.array(1)) == False`. The error is particularly annoying in conjunction with quantities, as `quantities.Quantity` is a subclass of `np.ndarray` and thus it's instances are not numpy array scalars.

Locate NumericalTrait(ndim=0) and the np.isscalar check. Reproduce the issue with np.array(1) and quantities.Quantity, then inspect the existing validation tests. Done means zero-dimensional arrays are accepted as scalars while nonzero-dimensional arrays remain rejected.

Written by the indexing model from the issue text.