Support for different intial guess modes

[tjgalvin]

The current approach to source finding with aegean is divided into two main stages -- one that create the initial estimate of the number of components (i.e. the definition of the model) and their estimated location with an island, and the second that performs the actual optimization of the model.

I am beginning to see cases where the optimized model is not entirely correct, and based on the residual of the fit it looks to me like the number of components is not correctly estimated. I think it might be worth considering including alternate methods of estimating the initial conditions of the model and island. It would also be good to be able to condition multiple models of the same island separately using different initial conditions and pick the best fitting one of the set at the end.