Minimize the numbers of file read in N simulation

[rosiereflo]

If we were able to provide the dyd/par files directly in memory for the N simulation we would go faster
It would also be interesting to have the same approach in SA but it is more complex (need to copy the objects and to add the contingency into the dyd model for each simulation )

[rosiereflo]

We need to make sure this is a bottleneck before going on