Implementation of mesh refinement

[SeverinDiederichs]

This issue serves as a notebook for the implementation of mesh refinement and should be updated regularly

First, a static mesh refinement based on Timon's original subgrid algorithm (https://doi.org/10.1109/AAC.2018.8659404) will be implemented.

A couple of technicalities need to be implemented:

• a grid for the refined level needs to be allocated (done in first preparation to enable multiple levels #523)
• IO for all levels (done in Mesh refinement: Adding IO #526)
• multiple Poisson solvers due to the different dimensions etc (done in Mesh refinement: multiple Poisson solvers #527)
• fix Poisson solvers on GPU (to be fixed in Mesh refinement: fix MR on GPUs #564)
• the box array of level 1 needs to be fixed (done in Mesh refinement: fix boxarray #529)
• the binning of the beam must be addressed. (this issue disappeared after the latest fixes)

Finally, some physics:

• the beam and plasma current must be deposited to the correct grids
• the value of the field on the main grid needs to be interpolated to the source term of the lower level grid
• the beam must be pushed on the lowest level, if applicable. (this is currently only stable, if the beam is fully contained on the refined level)

This should enable the most basic simulations, where only the transverse resolution is increased.
Next step should be the interpolation also longitudinally, to enable a refinement in all directions.

• enable longitudinally smaller patches than level 0. (to be done in Mesh refinement: enable level 1 to be longitudinally a different size than level 0 #546)
• enable longitudinal refinement. This requires: binning of the beam per level and a sub-loop over the longitudinal refinement ratio to solve multiple slices on level 1 for each slice on level 0. (to be added in Enable longitudinal refinement #574 )

Some cleaning:

• the IO box is much larger than what will be actually written. This should be adjusted. This will be solved partially in fix transverse size of IO for level 1 #569
• add CI

Known problems:

• plasma particles are not redistributed on the fine grid properly, if they leave the box due to ion motion, it crashes with a segfault (fixed in enforceBC works for all levels #566)
• slice IO seems to be broken with longitudinal refinement (fixed in Mesh refinement: fix slice IO #568)