Optionally reset velocities for MD

[ElliottKasoar]

Resolves #666

This is a slightly different fix to the discussion in the issue, to me it also seems like a reasonable solution that makes sense based on our current parameters (rescale_velocities in particular, since this already guarded an effectively duplicated self._reset_velocities() call):

• If rescale_velocities is True (default is False) then velocities are reset before heating and/or MD (assuming we're not restarting)
• If rescale_velocities is False, we don't reset the velocities, unless no velocities have been set, in the same way as NVT_CSVR. I'm also happy to leave the velocities at 0 if it makes more sense

I test that reading in a file with velocities (and rescale_velocities is False) keeps these velocities, so I think it addresses the main point of the issue.

There is also scope for debate about the additional resets we perform in between heating steps, and between heating and MD, but that could also be addressed separately.

How does this look/sound, @trdurrant?

[alinelena]

what I think is needed to have a test where we have an input file that contains velocities and without any rescale_velocities in the input we keep the user velocities

[ElliottKasoar]

what I think is needed to have a test where we have an input file that contains velocities and without any rescale_velocities in the input we keep the user velocities

Is that not what test_no_reset_velocities does?

lj-traj.xyz includes velocities, so I just reused it for the input file, and I check that with rescale_velocities=False (the default, but included to be explicit) the velocities are unchanged.

[ElliottKasoar]

what I think is needed to have a test where we have an input file that contains velocities and without any rescale_velocities in the input we keep the user velocities

Is that not what test_no_reset_velocities does?

lj-traj.xyz includes velocities, so I just reused it for the input file, and I check that with rescale_velocities=False (the default, but included to be explicit) the velocities are unchanged.

Actually the momenta are stored, not velocities, but I don't think this makes a practical difference.

[alinelena]

what I think is needed to have a test where we have an input file that contains velocities and without any rescale_velocities in the input we keep the user velocities

Is that not what test_no_reset_velocities does?

lj-traj.xyz includes velocities, so I just reused it for the input file, and I check that with rescale_velocities=False (the default, but included to be explicit) the velocities are unchanged.

cool!