I'm not sure how much sense this makes, but I am now explicitly dividing
the computed spectrum by dk, such that the integral of E(k)dk is the
total energy.

this emphasizes the transposition issue.

both NavierStokes and NSVorticityEquation fail with segfaults, but the
test code is minimally sane

In the past, if I discovered that the pp file was computed for different
iterations, I removed everything from it.
Which is a bad idea, hence I'm not deleting the datasets that I know
have to be deleted.

The particle datasets, in the case of a single particle cloud, were
being created with really small chunks.

this should fix part of the I/O issues.
And, as a matter of principle, we don't really need double precision for
these quantities, for the sort of statistics we want.

also, add the option of a regular cube as initial condition.