Move visualization functions to plotly and get rid of matplotlib dependency

[npatki]

Problem Description

Other libraries in the SDV ecosystem are using plotly to generate visuals. For example, the SDMetrics library offers visualization utilities using plotly.

However, the Copulas library still uses matplotlib. It would be nice to unify all visualizations to using plotly:

1. Plotly is interactive. You can zoom into specific areas, as well as toggle legend items on/off
2. We want to limit the number of external dependencies in the SDV ecosystem, as it makes it harder to maintain and forces us to make more frequent updates

Expected behavior

See Copulas Visualization Methods. Some directly correspond to SDMetrics so we can use the same graphing interface.

• copulas.visualization.compare_1d is equivalent to sdmetrics.utils.get_column_plot
• copulas.visualization.hist_1d can use the same visualization as above, but it only has 1 data series (color)
• copulas.visualization.compare_2d is equivalent to sdmetrics.utils.get_colunn_pair_plot
• copulas.visualization.scatter_2d can use the same visualization as above but it only has 1 data series (color)
• copulas.visualization.compare_3d can use the plotly 3D scatter plots
• copulas.visualization.scatter_3d can use the same 3D scatter plot as above
• copulas.visualization.side_by_side: This may be a candidate for deprecation. Since plotly is interactive, there is no need to pass in extra kwargs

Additional context

To keep things consistent, all visualization functions should also make the following changes:

1. Return the plotly.figure object instead of just showing it. If a user gets access to the object, they can make any modifications they like to the visualization
2. Use the same color scheme as the SDMetrics visualizations for consistency
3. Update any titles to accurately reflect what is being shown