Forest Carbon test data: invalid geometries after reprojection

[emlys]

Enabling gdal.UseExceptions made this issue apparent. Two tests (test_carbon_dup_output and test_carbon_full) started failing because of invalid geometries in the vector resulting from reprojecting the original model data. For #1570 I'm going to just skip over any invalid geometries, since that's effectively what was happening before, but we should check why these are being produced and if there's any way to avoid it.

[davemfish]

@emlys this sounds very familiar. Is it possible this was addressed here where James enabled parital reprojection, which I think skips over would-be invalid geometries? https://github.com/natcap/invest/pull/1646/files

[emlys]

@davemfish I confirmed I'm working on a branch that does have those changes, and the invalid geometries are not present before reprojection. Perhaps "partial reprojection" also means reprojecting just part of certain geometries (that are on the edge)? Just a guess