Feat/add-physical-validation

[thomasloux]

Summary

I've added tests to evaluate the correctness of integrators. I'm using physical_validation, used in gromacs test suite. The code is WIP, so that people can try it easily and update some parameters. In the future, the systems studied and the parameters will need to be fixed, or used the standard parameters. Note: this is important because the default parameters may be changed in an undesirable way.

What to do next?

• Check that the chosen parameters are correct default one.
• Run for longer to have convergence of the integrator.
• Choose the optimal pressure to correctly evaluate NPT integrators.
• Change default parameters for some integrators, most importantly for NPT NH.

Notes

• Parameters for NPT langevin are most likely wrong, as noted in Wrong unit of default b_tau in npt_langevin_init #453.
• C-Rescale Anisotropic does not seem to work, crystal general tend to grow too large. I haven't checked the trajectories, but this is something I've realized during development. Maybe we should add a warning when a user chooses NPT anisotropic for now.
• Langevin parameters need to be more studied, the cell volume variation is far from correct.
• I should add how to estimate the isothermal_compressibility for c-rescale, which is easy and fast using QHA and phonopy.
• I've already tested changes in Fix NPT Nose-Hoover things #520 and it works great. Actually I didn't even try without these changes, but I'm quite convinced the changes in Fix NPT Nose-Hoover things #520 are correct.