WIP: RegularSingleCoreDecomposition w/o ghost cells and comms

[RudolfWeeber]

Description of changes:
this is an early sketch for a cell system based on RegularDecomposition, but for simulations on a singel mpi rank. It does not have ghost cells and ghost communication. Cells are directly connected acrsoss the periodic boundary.

Performance implications

• A system with 1000 particles at sigma=1 and volume fraciotn=0.1 has box size of abou 17**3.
• For electrostatics with a cutoff of a bit over 2, this would make 83 cells without ghost cells but 103 with ghost cells.
• 83/103 ~= 0.5
• Ghost cells, even ona a simulation with a single core, are accessed twice per time step: once to update ghost positions and once for the ghost force reduction. Also, they use up memory and memory bandwidth, so that only a smaller system fits into the cache.

Current status:

• Really just a sketch. doesn't compile and no Python interface

Technical notes

• RegularDecomposition uses a ghost_cell_grid wher index 0 and n refer to the ghost layers and the stuff in between (1 to n-1) are the local cells. The singule mpi rank variant will only have a cell_grid with local cells running from 0 to n-1.
• We'll see if other parts of Es make assumptions about ghost cells being present.

Outlook

This may potentially be useful in conjunction with shared-memory parallelism and an OpenMP enabled fftw for P3M.

[espresso-ci]

Your pull request does not meet our code documentation rules. The Doxygen documentation generated 1 unique warnings (total: 15, ignored: 12):

src/core/cell_system/RegularSingleCoreDecomposition.hpp:
  line 195: Reached end of file while still inside a ()

You can generate these warnings with make -t; make doxygen; ../maintainer/CI/dox_warnings.sh using the maxset config. This is the same command that I have executed to generate the log above.