Corrected stimulus preparation and update for cosimulator.py

[dionperd]

I moved the _prepare_stimulus() method call to the configuration of the CoSimulator, because during co-simulation the simulation_length is modified and equalized with the synchronization_time, leading to an erroneous reconfiguration of the stimulus in time.

I added an _current_stimulus property to the CoSimulator, in order to keep the stimulus between calles to call() method.

I also overwrote the _loop_update_stimulus method to use

stim_step=step-1

instead of stim_step=step-(current_step+1)

because current_step is updated evey synchronization time at each call().

[liadomide]

I think PR #673 should be merged before this one, rebase against master, and then contrib unit-tests should be passing

[maedoc]

hi @dionperd, I'm a bit late for this one. still relevant for merging? will you rebase as Lia suggested?