implement weighted aggregate operations

[gdementen]

here is crude code I wrote for mean for my own uses:

The interesting part is that it allows weights to have more dimensions that the original array, which is sometimes useful.

    def mean(self, *args, **kwargs):
        keepaxes = kwargs.pop('keepaxes', False)
        skipna = kwargs.pop('skipna', None)
        weights = kwargs.pop('weights', None)
        if skipna is None:
            skipna = True
        if weights is None:
            func = np.nanmean if skipna else np.mean
            return self._aggregate(func, args, kwargs, keepaxes=keepaxes,
                                   commutative=True)
        else:
            func = np.nansum if skipna else np.sum
            if np.isscalar(weights):
                # assume weights is a label along an axis
                weights = self[weights]
            wsum = (self * weights)._aggregate(func, args, kwargs, keepaxes=keepaxes, commutative=True)
            expanded_weights = weights.expand(self.axes)
            num = expanded_weights._aggregate(func, args, kwargs, keepaxes=keepaxes, commutative=True)
            return wsum / num