Try better plasma particle ordering in CD

[MaxThevenet]

This commit proposes a few combination of (i) particle ordering at initialisation and (ii) deposition scheme (1 thread per particle, 1 thread per particle per x cell, 1 thread per particle per x cell per y cell). For the moment, only the best combination so far will be merged, but it is worth looking into this again to check if there are even better combination. In particular 1 thread per particle per y cell was not tested, along with initialising particles per 4x4 group of cells.