this time without looping over the whole grid. the limited rates are
only losely coupled to be the ones used by the actual source terms, so
if the rate limiting code is changed the statistics gathering piece
must also be adapted. This is kind-of sub-optimal because it requires
some code duplication, but the alternative would be _much_ less
performant...
during debugging this was used to make sure that the rate limits are
correct. the issue is that this is slow if many wells are involved
because it uses a full iteration over the grid for each well to
calculate these rates...
