int32 overflow for particle checkpointing fixed

[Alcauchy]

• If a simulation has a large amount of particles, npart>2^32-1 (e.g., for a 40 000 x 40 000 simulation with ppc=16) would lead to overflow in nout_total, nout_offset, npart_total, npart_offset using npart_t during checkpoint writing, corrupting checkpoint data (see screenshot of density field attached).
• This is fixed by declaring the variables mentioned above as std::size_t at particle_io.cpp.
• The fix was tested on frontier with 40000x40000 simulation, and now the data is saved correctly.