RunModel_LLR bug in final state creation

The modification 31f8e94e try to take care of the length of IndPeriod_Run in order to avoid an overflow when picking final states.

This modification doesn't take into account the availability of data in the warm-up period, this latter should be used.

The main restriction with the current implementation is for example in the case of a step by step run of the model with a delay greater than one time step where most of the information will be lost. Moreover, at the next run, the use of this state vector will crash the function because the length of initial state vectors should be exactly the same as the length induced by the delay.