Configurable MNL coefficient bounds

[smmaurer]

In the MNL estimation code there are bounds set on the coefficient values, i guess so that unstable models will converge on something. The bounds are set at (-3, 3) in the older UrbanSim codebase, and (-1000, 1000) in ChoiceModels.

https://github.com/UDST/urbansim/blob/master/urbansim/urbanchoice/mnl.py#L178
https://github.com/UDST/choicemodels/blob/master/choicemodels/mnl.py#L589
https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.fmin_l_bfgs_b.html

We should make this configurable from a higher level, like in the model template estimation settings.