Bring None or NaN values to the back

[hadim]

It might be nice to always or behind a flag bring None or NaN values to the back in the Z axis of the scatter plot.

One way to do that is to perform a sort operation, such as:

data = data.sort_values("x", na_position="first")

It works well but comes at the cost of a sort operation and won't work when the color_by value is modified dynamically.

I am not familiar with the API and the JS code, but maybe there is a way to efficiently modify the z for None or NaN values?

[flekschas]

To avoid having to re-send, re-index, and re-upload data when one dynamically changes a setting that contains NA (e.g., scatter.color(by='column_with_nans')), this feature first requires the ability to tell regl-scatterplot in which order to draw points (flekschas/regl-scatterplot#162). Fortunately this should be fairly easy to implement.

API-wise I'm thinking of something like:

scatter.data(sort_records_with_na="first" | "last" | None)