Implement simple prediction of non-existing modes

[doorleyr]

Implement prediction of new modes, starting with (i) floating micro-mobility and (ii) an autonomous shuttle service. This can be done by fitting a quasi-nested logit model as described by @wangcan1213 in this notebook: https://github.com/wangcan1213/ipynb/blob/master/mocho_logit.ipynb
Micro-mobility can be assumed to be similar to cycling and autonomous shuttle similar to public transit.

When the model is fitted, we can include these modes in the simulation using simple assumptions for the generalised costs of the new modes. As a next step, we may refine these generalised costs using simulation.