CR3BP: Clarify computation of manifolds for libration orbits
After the correction of Hipparchus 253 related to Eigen decomposition, failing test cases appeared in Orekit (see #1205 (closed)).
These tests were fixed so that Orekit behavior stayed the same as before.
However, it appeared we couldn't clearly explain how the manifolds are obtained in LibrationOrbit:
- Unstable manifolds use the real part of the 1st Eigen vector of state transition matrix
- While stable manifolds use the real part of the 2nd Eigen vector of the same matrix
This computation needs references and/or explanations on this choice of eigen vectors.