Fix spurious copy() statement in LAMMPS interface

[RandomDefaultUser]

As discovered while implementing #226 the current LAMMPS interface makes a copy() statement necessary for issues related to the underlying numpy routines. See #226 (comment), #226 (comment) and #226 (comment). This is not a problem for now, but it might become restrictive once we really push the limits system sizes MALA can treat.

[RandomDefaultUser]

With this bug fixed numpy/numpy#18294, it would be wise to re-investigate this.

[RandomDefaultUser]

Since we are moving to FP32 data, which means that every LAMMPS call will (by default) include a cast to FP32, the copy is no longer of any concern. It will only rarely be called at any rate. Thus, this issue will be fixed by #435 .