Should put a parameter table somewhere for the pre-defined cosmologies for reference

If we separate out the implementation of dark energy into other types (e.g., cosmological constant, w0wa, etc) we could encapsulate them in the cosmology objects which could allow more efficient dispatches (e.g., H for a cosmological constant is simpler than for w0wa).

This may complicate some other calculations so worth thinking more about what interface we would want to enforce.

Should probably try to change nu_relative_density to be smarter about the number of massless neutrinos if the length of the m_nu iterable is not the same as N_nu

Powers of Unitful quantities (like 1 / Unitful.Mpc^3) print to console differently on Linux and on Mac OS, causing our doctests (which require an exact print match) to fail. Refactor the doctests with Unitful outputs to do a \approx check and return true instead. Not as nice to read, but should be more stable long-term, as it doesn't depend on how Unitful elects to print quantities to the terminal.