Parallel mesh assignment and simplification of M2L kernel
Adds:
- List of purely local cells.
- Use the threadpool to assign the gpart to the mesh.
- Simplify the dipole terms in the M2L calculation.
- Make a symmetric version of the M2L kernel.
- Use an approximation of erfc to avoid external function calls.
- Add the P-Millennium-384 example.
- Add the P-Millennium-768 example.
Edited by Matthieu Schaller