checkpoint is somewhat improper because checkpointing functionality was
implemented in vorticity_equation instead of
direct_numerical_simulation, so that would be the next step.

preliminary test works with no obvious problems.

now code compiles

not tested yet.

there were a few issues with read_parameters, which is now a virtual
function in NSVE.

this means i defined a virtual "step" method, as well as "do_stats" and
"write_checkpoint".
I think this makes it even more clear what job is assigned to which
class.

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