Compute mask sizes more carefully in DensityFiltration

[ulupo]

Currently, the logic in

giotto-tda/gtda/images/filtrations.py

Lines 1166 to 1169 in d089002

	self._size = int(np.ceil(
	pairwise_distances([[0]], [[self.radius]], metric=self.metric,
	**self.metric_params)
	))

assumes a fair amount form the metric used. In particular, it assumes that the metric has a counterpart in 1D and that a cubical "container" for the appropriate mask can be found purely by considering this 1D counterpart.