Add dynamic carbon cycle

[hdrake]

At some point, we should replace our simple zero-dimensional transient climate response model with a two-box models for surface ocean / atmosphere and deep ocean (e.g. Held 2010). Similar for adding a two-box carbon cycle for a ocean carbon sink (e.g. Glotter 2014).

[hdrake]

Implemented a version of the Held two-box energy balance model in https://github.com/hdrake/OptimizeClimate/tree/two-box-temperature

[hdrake]

As of a1ab9f5, the master branch now includes a two-box model. Dynamic carbon cycle may be implemented further down the line, but it is mathematically and conceptually much more elusive (to the general user) than the rest of the model, so it is unclear if the benefit of greater quantitative accuracy is worth the added complexity.