geo-shape-doc-values centroid calculation should handle polygons as lines/points and lines as points

[talevy]

• Polygons with no area should not be treated as polygons — they should be treated as lines.
• Lines with no length should not be treated as lines — they should be treated as points.

possible solution A

This can be achieved by updating the DimensionalShapeType of the geometry for such geometries. The downside of this is that then we would love information about what type of geometry the field value was.

possible solution B

The other option is to keep the original DimensionalShapeType, and instead have a place-holder weight that is close to 0 for such geometries.

relates to #50834

[elasticmachine]

Pinging @elastic/es-analytics-geo (:Analytics/Geo)

[talevy]

@nknize and @iverase, do you have a preference for how this issue is to be solved? Currently such values return NaN for longitude and latitude. At the time, I had tests for this and thought it was OK, but now I realize such geometries are very much expected in the wild.

[iverase]

My opinion here is those geometries are degenerated and they should be treated as the final shape. So if a polygon becomes a line, the DimensionalShapeType should be change accordingly. I think it is more important the current geometry than the original geometry type.

[talevy]

this was closed in #52500