.value was deprecated, so I now use [...], which should work for old
enough versions (they recommend [()] instead).

I know the arithmetic mean thing is more polite, but I'm going to use
the old version, so that the test still passes as is.
And everyone does it this way anyway...

not tested yet.

keep NSVEparticles, since it's a child of NSVE and everything is more
consistent this way

results should (and seem to be) equal within numerical precision (maybe
even binary for good settings).

I need to figure out how to define the read_parameters method properly