mpol.plot.vis_histogram_fig to accept loose visibilities?

[iancze]

Is your feature request related to a problem or opportunity? Please describe. The new mpol.plot.vis_histogram_fig routine is great! But it requires a mpol.datasets.GriddedDataset object as input, which requires the user to make assumptions about the image size via a GridCoords object so that they can use the DataAverager to produce a GriddedDataset. They're probably going to want to do this at somepoint anyway, but it seems a bit premature to make them do this while they're still familiarising themselves with the visibility dataset.

The current docs also have some unit=:math:[klambda] that didn't render.

Describe the solution you'd like Could vis_histogram_fig accept (non-Hermitian pairs) uu and vv numpy arrays directly, as DataAverager or DirtyImager do? Could also borrow logic from inside DirtyImager to expand the arrays so that they properly include Hermitian pairs in the histogram.

And then the logic inside vis_histogram_fig could be simplified slightly since we don't need to worry about indexing zero or non-zero GriddedDataset cells.