bug in rdf/gran

Hi,
Greeting,
firstly, I did not see files that looks like 'compute_rdf_gran.h and compute_rdf_gran.cpp' in the released LIGGGHTS version,
so I just created them because they are very similar to compute_rdf.h and compute_rdf.cpp. But still have bugs.
1. It seems the skin of neighbor bin is not accounted for before computing the width of the shell, though it is addressed in the manual.
2. It is safer to recount the number of particles before start calculate the g(r) for granular, since number change in the simulation box is often in granular simulation, for example, pouring.
The fixed files are attached.
Hope it is helpful.