Refactor and introduce SphericalOverdensity model

[mabruzzo]

This PR lightly refactors the codebase so that we now define a very simplistic SphericalOverdensity model.

The main thing that changes here are that:

• the Header::sphere_radius, Header::sphere_*, data members are now tracked in a dedicated model object
• it makes sense for this to be a model object because the model properties (e.g. overdensity, bkg_density, center-position, initial_radius) must be known for both initialization AND for setting boundaries
• the associated overdensity and bkg_density are now runtime arguments (the model machinery requires each model to be configurable by one or more runtime parameter -- I picked these somewhat arbitrarily). I added documentation for these new parameters
• as part of this PR, I ported over the documentation for the init runtime parameter

Motivation: This PR, #524 and #525 will all be required in order to cleanup the Grav3D class. Plus, this is a much more simple example of the kinds of things that should be implemented as model objects.

Note

I was very careful to try to avoid breaking things. However, I'm pretty sure that test-problems leveraging this machinery are fundamentally broken. Unless I'm totally off-base, this requires non-periodic boundaries (if we want to run it with periodic boundaries, then we can cull lots of things -- this doesn't even need to be a model). Since SOR doesn't work, I think we're run this with Periodic Paris. But, we'll see as I continue trying to clean things up