sphdistance can miss some nodes

See the [forum](https://forum.generic-mapping-tools.org/t/question-about-sphdistance/1096/2) issue for specifics. Looking at the nodes that are not set, they look like this:


I think the reason this happens is that Sabin has lots of very acute spherical triangles from his densely sampled points along the ridge to a far point on another ridge. Due to some unclear roundoff the test that checks if a node is inside may somehow fail nd leave these behind.

One solution I can implement is to loop over the grid, find these nodes and replace them with the 4-star average.