Incorrect virials?

[janosh]

while looking for possible reasons for slow FIRE convergence reported by @AdeeshKolluru, i noticed a handful of places where the virial appears to be incorrectly calculated in both integrators.py and optimizers.py:

https://github.com/Radical-AI/torch-sim/blob/95bba41eb7e7d23ebecec9a74646cab2e91c0987/torch_sim/optimizers.py#L318

https://github.com/Radical-AI/torch-sim/blob/95bba41eb7e7d23ebecec9a74646cab2e91c0987/torch_sim/optimizers.py#L423

https://github.com/Radical-AI/torch-sim/blob/95bba41eb7e7d23ebecec9a74646cab2e91c0987/torch_sim/optimizers.py#L893

https://github.com/Radical-AI/torch-sim/blob/95bba41eb7e7d23ebecec9a74646cab2e91c0987/torch_sim/optimizers.py#L1025

https://github.com/Radical-AI/torch-sim/blob/95bba41eb7e7d23ebecec9a74646cab2e91c0987/torch_sim/optimizers.py#L1321

stress and pressure both have units of force per area so need to be added first before multiplying with volume to get an energy. @abhijeetgangan this should be -volumes * (stress + pressure), no?

https://github.com/Radical-AI/torch-sim/blob/95bba41eb7e7d23ebecec9a74646cab2e91c0987/torch_sim/integrators.py#L606

here it should be -volumes * (state.stress + pressure_tensor)