Add grid subdivision / rasterization algorithms

[dbaston]

As discussed in #967, adds functions for:
(1) subdividing a geometry by a rectangular grid
(2) computing the fraction of each cell in a grid that is covered by a polygon

Performance in the included benchmark is 1-3 orders of magnitude faster than repeated calls to Geometry::intersection

Code was adapted and pared down from https://github.com/isciences/exactextract

[pramsey]

This looks very exciting, I'm fine w/ the CAPI and not qualified to judge the implementation. The tests look good, maybe see if any of the codecov suggestions for complete coverage are easy to knock off.

[dbaston]

Still working through a couple of minor issues with this. I think it could go in the beta, I'm not seeing anything I can't resolve before release.

[dbaston]

Current status: this produces valid polygons in all cases that I've tested, but the polygons do not necessarily form a valid coverage if polygon boundaries follow grid cell boundaries. Here is an example:

Subdivided to produce:

In this case, some of the fully interior polygons are generated as squares without redundant nodes along their edges. As a result, the polygons do not form a valid coverage. Because the missing nodes may only be on a horizontal or vertical edge, they do not cause an overlap or gap.

[theroggy]

In this case, some of the fully interior polygons are generated as squares without redundant nodes along their edges. As a result, the polygons do not form a valid coverage.

Because the missing nodes may only be on a horizontal or vertical edge, they do not cause an overlap or gap.

It isn't really clear to me where the missing nodes are, but this last sentence is an important nuance indeed... because of this overlay operations should not cause slivers,... so at least for my use cases, it not being a valid coverage is not a problem.

[dbaston]

It isn't really clear to me where the missing nodes are,

I mean that polygon "A" should have a node at the location pointed to by the arrow. If it does not have a node at this location there will be no gaps or overlaps between adjacent polygons, but the polygons will still not be considered to form a valid coverage.

Either way, I believe all issues in this category have now been addressed.