add calc_force_and_virial

[ToruNiina]

after adding virial calculation, the performance decreases ~15% in some cases. Maybe it is better to control the virial calculation, by adding calc_force_virial, calc_force_energy_virial, calc_force.

The implementation should be shared among those, by adding some template parameter to use constexpr if, as suggested in #291 .

[ToruNiina]

done by #304 .