Gadget-2 particles memory footprint

Particles could be made a tiny bit leaner by doing the following:

• Move div_v to the density sub-structure.
• Replace curl_v in the force sub-structure by the Balsara term and compute it in the hydro_prepare_force() routine.
• Move entropy_dt to the force sub-structure.
• Move a_hydro and h_dt to the force sub-structure. Although that might have consequence on the other schemes.
• Replace pressure with P_over_rho which then has to be computed in the hydro_predict_extra() routine.

That last one is not a space optimisation but a speed optimisation.

Doing this implies modifying critical part of the physics implementation as well as the corresponding unit tests.