Don't assume all flat polygons face up

[uclaros]

Description

QgsTessellator uses a shortcut and considers all 2D polygons and 3D flat (all vertices having the same Z) polygons to be facing up, returning a positive normal and counter clockwise triangles, regardless of the original geometry's winding order. This was also unaffected if the invertNormals parameter was set to true, leading to more confusion.

With this PR, the flat polygons' winding order is taken into account, producing counter clockwise triangles and positive normals for counter clockwise polygons, and clockwise triangles and negative normals for clockwise polygons.
If invertNormals == true then both the normals and the triangle winding order is reversed.

Edit: However, there are polygons that are not made to be rendered in 3d but rather are just used as footprints for extrusion. Their data provider might even force a winding order, like the buildings shapefile in the test suite. So when we have flat polygons that are extruded, we always treat them as up-facing, regardless of their winding order or the invertNormals parameter.

Fixes: #52832

[github-actions]

🪟 Windows builds

Download Windows builds of this PR for testing.
Debug symbols for this build are available here.
(Built from commit 944c08d)

🪟 Windows Qt6 builds

Download Windows Qt6 builds of this PR for testing.
(Built from commit 944c08d)

[nyalldawson]

Can you explain why these have changed? The tesselator used in QgsVectorLayerProfileGenerator is set not to calculate normals...

[uclaros]

The normals are always calculated and used to define the winding order of the resulting triangles. The tessellator addNormals parameter sets whether the normals will be included in the result data, which in this case are not needed.
Still, I'm not exactly sure why the tests needed adjustment. It seems there were some rounding errors caused by the triangles being calculated in the opposite orientation.