[WIP] Globe axis

[asinghvi17]

This PR builds towards a 3D Globe Axis, that currently will accept a Geodesy.jl datum as the ellipsoid definition, and try to work in ECEF coordinates. We can add more functionality, integration with CoordRefSystems.jl datums, etc later.

This axis will:

• Automatically apply a transform_func to all plots, going to 3D cartesian space.
• Lock certain zoom keys so you can interact with this like Google Earth
• Add a "satellite" camera-interaction API so you can move the camera in spherical space.
• Subsample polygon meshes so that they always lie on the sphere to some approximation, not under it
• Allow the following GlobeAxis-specific keywords:
  • zlevel: set the "z-level" i.e., some radius offset, so it's easier to layer plots (e.g earth image below, polygons above, etc
  • source: same as GeoAxis source, indicate a source CRS
• Add a "background" sphere by default, using a modified icosphere algorithm, that could at some point be parameterized to hold some background image as well.

More todos to come. This is an expansion of the Geodesy.jl approach that was taken previously

[lupemba]

This is really awesome. Do you think it will work with Tyler.jl?

[asinghvi17]

@lupemba I'm going to be working on that sometime this month :D - just got Tyler to support GeoMakie (though interactivity is a bit buggy).

The challenge here is that you have to figure out the intersection of the camera's view frustum (essentially a "pyramid section" along the central axis) with the sphere, and then figure out which lat/long area is visible. This is solveable, of course, but tricky, so it will be some time until I can figure out exactly how to solve that. Once that is done, then Tyler will work with this, but will have "holes" at the poles beyond -85 and +85 degrees latitude. That's an inherent limitation of the Web Mercator crs, so we'll just have to deal with it I suspect.

[asinghvi17]

I'm going to merge this anyway so we get the new Makie compat, and marked the GlobeAxis still experimental. It's going to change a bit at least in the interim anyway.

Datum shifts are still not implemented there and neither are spherical geometry or dynamic polygon resampling in meshes.