Add LinkedGeo <-> GeoMultiPolygon conversion functions

[ajfriend]

• Add internal functions linkedGeoPolygonToGeoMultiPolygon and geoMultiPolygonToLinkedGeoPolygon for converting between the two polygon representations
• Both directions validate geometry: every loop must have >= 3 vertices and every polygon must have an outer loop (E_FAILED otherwise). Empty multi-polygons (0 polygons) are allowed.
• Changes should have 100% line and branch coverage.
• Part of Replace cellsToLinkedMultiPolygon implementation with cellsToMultiPolygon #1127

[coveralls]

coverage: 99.095% (+0.08%) from 99.013%
when pulling b9f33ed on ajfriend:aj/poly_convert
into 8085979 on uber:master.

[isaacbrodsky]

I would err on the side of keeping these as internal functions rather than exposed, but as they would not be required for the bindings maybe it doesn't matter too much?

Edit: we can of course add fuzzing targets even for internal functions :)

[ajfriend]

I would err on the side of keeping these as internal functions rather than exposed, but as they would not be required for the bindings maybe it doesn't matter too much?

I agree that we don't need to be as strict about these conversion functions: 1) They wouldn't be exposed to users of the bindings. 2) These are really only for backwards compatibility, and we'll be recommending that people use cellsToMultiPolygon going forward. So, ideally, users and bindings would never need to touch (or convert to/from) LinkedGeoPolygon.

That being said, if a user does need to convert between them, I think providing the utilities to do so is helpful, rather than having them rewrite it themselves (like we historically did in the bindings).

[ajfriend]

Agreed with @isaacbrodsky that we can have these functions be internal for now. It would be easy to expose them in the future if desired.

[ajfriend]

Updated based on comments, and code coverage stays at 100% for changes.

[justinhwang]

lgtm!