Improve CrossCorrelation memory consumption and speed

#5
Calculating Cross-correlation of Bold time series.
It took 2,5h to calculate the cross-correlation of a 10s Bold time series. We canceled the calculation after 8h for a 40s Bold time series.
...follow up on this:
We noticed the follwoing: two more pictures of the "activity report" are attached.
First one after 40s Bold simulation. Second during calculation of CrossCorrelation. The amount of python processes and also the amount of RAM used for it is increasing step by step. Could this be the reason for the problem ?

#5. We are using numpy.corrcoef() for every state-variable & mode of the resulting TimeSeries.
Right now, we are keeping in memory all the intermediate results, until the full operation is finished, and then we dump it on the disk.
We could improve a little and write after each state-variable & mode result. I added a task for this.