Wolf own solver

There is some deprecated code in folder solver_suitesparse that should be either updated or removed.

Maybe we should decide if we provide a simple solver along with wolf or we force the dependency on ceres (or GTSAM eventually).

added Code organization Idea labels

mentioned in issue #63 (closed)

@mederic_fourmy is #295 only for obtaining the Hessian or also solving?

The initial goal was to compute the hessian yes. Since the weighted jacobian matrix is readily available, it is easy to compute the hessian by just returning A.T * A. In the branch related to #295 I adapted the solver to nonlinear states localparameterization and to the new functions of SolverManager that were not here at the time. However, a quick test on some IMU gtests I could not attain convergence and I dropped it for now.