[BUG] Point-in-polygon Primitive Should be Consistent Over Boundary Condition

[isVoid]

Describe the bug
Consider two axis aligned rectangle:

[(-1.0, -1.0), (0.0, -1.0), (0.0, 1.0), (-1.0, 1.0)]
[(0.0, -1.0), (1.0, -1.0), (1.0, 1.0), (0.0, 1.0)]

and a test point (0, 0).

This point is on the edge of these two rectangles. Is this point in the rectangle? The answer should only be always yes, or should be always no for both cases. Not yes for one and no for another.

Not yes for one and no for another.
This is what's happening in current PIP kernel.

Steps/Code to reproduce bug
See test:

cuspatial/cpp/tests/experimental/spatial/point_in_polygon_test.cu

Line 157 in 930a2f3

TYPED_TEST(PointInPolygonTest, EdgesOfSquare)

Expected behavior
Result should either be b0000 or b1111. Better still, user definable on boundary conditions.

[thomcom]

I think that #750 fixes this.

[harrism]

No, the crossingsMultiply algorithm is fundamentally inconsistent. Working on it.