new model: Hyperbolic with non-linear utility function

[drbenvincent]

Add a new model, hyperbolic discounting but with a power-law nonlinear utility function. This will mean that this toolbox is no longer strictly focussed on discounting-only models, but becomes a more general decision making toolbox. In this model, present subjective value is calculated as

V = u(reward) * 1./(1+k*delay)

where

u(reward) = reward^a, if reward>0

and

u(reward) = -lambda(-reward)^a, if reward<0

where lambda =1 in order to keep this model very simple, ie no loss aversion parameter. We also only have one exponent parameter which is applied to gains and losses. These simplifications can be lifted in a future, model with more parameters.

• implement separate model + add to test suite
• implement mixed model + add to test suite
• implement hierarchical model + add to test suite
• update discount function plotting. Because discounting is now applied to u(reward) not just reward, I don't believe it is meaningful any more to plot data points on the discount function plots. Because it is not the raw reward values that are being discounted with time any more.
• update utility plots. Add in a new subjective utility plot to visualise the subjective utility function
• test the model's ability to account for magnitude effect data. Possibly using as the basis for some useful documentation