Optimize the dynamical fitting!

[3fon3fonov]

I found that all BS dynamical models call some unnecessary subroutines. Overall I managed to reach ~ 10 % speedup when I comment out these calls. However, I need to restructure a bit the dynamical codes, which will be done when I translate everything with f2py.

This is just a reminder to make the dynamical code run faster.

[3fon3fonov]

I could not resist and I did some of the optimizations. However, I found that the mixed keplerian-dynamical fitting routine is super inefficient w.r.t. the dynamical-simplex. This is most likely because it uses some parts of the variational equations code (see Tan et al. 2013), but we do not need the derivatives in the Simplex optimization! I cannot remember why was this coded that way, perhaps this was overlooked..... So there is still some work to be done on the RV dynamical models.

[3fon3fonov]

This is more or less done. Closing.