Issues can appear in orbit determination when selecting the central attraction coefficient (µ) as a parameter to estimate.
In the specific case where the user selects µ as a parameter to estimate but does not explicitly add an attraction force model to the numerical propagator builder.
The propagation is still correct as a simple Newtonian attraction force is automatically added in the numerical propagator.
However the µ parameter selection is not known by the propagator.
This results in wrong computation in the Jacobian of the model.
The bug is silent. Everything works fine... but the results are wrong.

Proposed fix:
On the model of what is done in class NumericalPropagator, automatically add a Newtonian attraction force in the class NumericalPropagatorBuilder if no attraction force has been added yet.
Do this right before the creation of a propagator (in method NumericalPropagatorBuilder#createpropagator)
Thus the selection of the parameter µ will be known both by the builder and the propagator.

(from redmine: issue id 359, created on 2017-08-29)