Es walberla

[itischler]

fixes espressomd#4028

[RudolfWeeber]

Thank you. I just tried merging this PR.
However, thee are still some issues:

• the thermalized tests (such as lb_thermostat.py and lb_stats.py) return Nan.
• There is quite a lot of code duplication, because the eentire constructoin path exists twice, once for relaxation_rates and once for (shear_viscosity, magic_number).
• You use the same unit conversion for viscosity and relaxation rates in lb.pyx. I'm not sure that's correct.

Could you please centralize the viscosity to relaxation rate (and back) conversion in functions, e.g., in src/walberla_bridge/lb_relaxation_rates.hpp?
I.e.,

RelaxationRates relaxation_rates_from_shear_viscosity(double shear_viscosity, double magic_number=3./16.);

and also the shear viscosity from relaxation rates.

You could than expose this to Python, do any conversion of user input there, and have the entire construction path just once with RelaxationRates as argument.
Furthermore, you can make use of this in the viscosity setter/getter of the lb models.

I deferred merging for now, particularly, since the lb_stats.py and lb_thermostat.py tests are broken. Could you please take an other look?

(FYI: @jngrad)