Pass time step and temperature as function arguments

[jngrad]

Fixes #4247, partial fix for #2628

Description of changes:

• reduce usage of the temperature and time_step globals
• combine the int fluidstep and double fluidstep static globals into a single static global
• only increment LB RNG during LB integration (changes the thermalization random sequence when tau != time_step)
• reduce code duplication, replace raw pointers by references, improve separation of concerns

[RudolfWeeber]

I think it’s unit conversion. dt^2/agrid is the conversion factor for a force, I guess. The particle force is stored in MD units on the GPU, so the conversion from LB units has to happen somewhere. But the details of LBGPU-uints are a mystery to me. This will only get disentangled with the switch to Walberla, I guess.

[KaiSzuttor]

I think it’s unit conversion. dt^2/agrid is the conversion factor for a force, I guess. The particle force is stored in MD units on the GPU, so the conversion from LB units has to happen somewhere. But the details of LBGPU-uints are a mystery to me. This will only get disentangled with the switch to Walberla, I guess.

ah I thought that I once had removed almost all MD units. But probably only from the CPU version.