fix: Improve handling of auto-fitting with many features and the antimeridian on geo traces

[camdecoster]

Description

Improve handling of geo trace auto-fitting with multiple features and geometry that crosses the antimeridian.

Closes #7943.

Changes

• Refactor geo traces to take all feature geometry into account when determining the fit range
• Add tests for plots involving the antimeridian
• Add tests for new helpers

Screenshots

Before	After

Testing

• Be on v4.0
• Load Plotly devtools
• Enter the following into the browser console:

    Plotly.newPlot(gd, [{
        type: 'choropleth',
        locations: ['CAN', 'USA', 'MEX'],
        locationmode: 'ISO-3',
        z: [3, 2, 1],
        showscale: false
    }], {
        geo: { projection: { type: 'equirectangular' } },
        width: 700, height: 450
    });

• Note that the plot is split across the antimeridian
• Switch to this branch
• Run that snippet again
• Note the the plot is fitted properly

Notes

• This change is necessary is because geometry that didn't cross the antimeridian wasn't being normalized in the same way that geometry that did cross the antimeridian was
• I reworked the way the range is padded as well to better match what happens in getAutoRange
• A number of geo trace helpers were eliminated/changed/added to support this new method of fitting

[emilykl]

I don't like this pattern of describing what the calling function will do. The docstring should document this function's behavior, not the behavior of the calling function.

Suggested change

	* `fitGeojsonCoords` is empty — in which case the caller falls back to the data.
	* `fitGeojsonCoords` is empty.

[emilykl]

From a data-visualization perspective, probably better to turn off auto-fitting for this mock.

[emilykl]

Nice