Oep

I changed the following: (a) in the first iteration only kli is used, this avoids the wrong initialization. However this change only matters, if the vhxc is not loaded from restart file as is recommended. (b) the oep test in the testsuite now reproduces the result in the paper and should also run faster. (c) oep%scftol%final_tol can be used to control accuracy of the solver. Further, I plan to keep this branch and may change some stuff later.