Need a spherical bounds equivalent to PlanarBounds

[robertkleffner]

Right now each projection describes a range of x,y values that are valid for inverse projection into the spherical space.

Most projections have the same valid projection domain: the whole sphere. However, special case projections like Gnomonic and some conical projections are only valid for a subset of the sphere. And many whole sphere projections have odd behaviors or diverge at certain limits points, such as the poles or the prime meridian.

Consider adding a SphericalBounds method to describe the inputs, or at least a IsValid(lat,lon) type method which we can use to help filter tests and provide a better way to document these cases to end users.