add poisson solver to generate electrical field from non neutral charged gas densities

[psychocoderHPC]

Provide poisson solver

• add unit test
• move PIConGPU bug fix to independent PRs
  • main.cpp simulation lifetime
  • PMacc reduce buffer init
• move forEachCell to PMacc
• generalize stencil() and move it to PMacc (low priority)
• documentation (this is a task for a physicist)
• squash commits

The poison solver is only solving the E-Field for the initial gas density not for densities created during the slide!

Use picongpu --help to see how to enable the poison solver

Poisson solver:
  --poisson.activate                    enable poisson solver
  --poisson.maxSteps arg (=2000)        maximum number of steps for the 
                                        preconditioner
  --poisson.epsilon arg (=1e-08)        maximal allowed error of the poisson 
                                        solver
  --poisson.preconditioner.disable      disable poisson solver preconditioner
  --poisson.preconditioner.maxSteps arg (=20)
                                        maximum number of steps for the 
                                        preconditioner

output during execution

...
PIConGPUVerbose PHYSICS(1) | Poisson solver:
PIConGPUVerbose PHYSICS(1) |   - converged after 74/100 iterations with norm=509.216, total epsilon=5.15989e-09
PIConGPUVerbose PHYSICS(1) |   - duration 3.5903 sec
...

[PrometheusPi]

@BeyondEspresso Did you already have time to test this (as discussed in the last dev meeting)?