Fix modular simulator MTTK
The modular simulator MTTK pressure coupling was slightly off due to
different reasons:
* The PV term was taken at the time of the update of the extended
degree of freedom rather than at the time of reporting. This leads to
slight differences when using nsttcoupling and nstpcoupling > 1.
* The NH chain coupled to the thermostat would alter the extended
variable and hence the integral before it is reported. This NHC update
already belongs to the next step, so shouldn't change the reported
integral.
* The NH chain couple to the thermostat would not update the propagator
scaling. Again, this leads to very slight differences when using
nsttcoupling > 1.
These slight inaccuracies went undetected because they are minimal in
normal usage. To emphasize them in testing, the compressibility in the
simulator equivalence tests was strongly increased. This shows the
differences and would make the tests fail without the fixes in the
current change.
To implement these changes, the responsibility to calculate the
integral and set the scaling factor was moved from the elements to the
MTTK data structure which is accessed by all elements changing the
external degree of freedom and the box.