Apparent (critical) bug in Pythia8_i

This fixes a critical bug in Pythia8_i that results in a memory allocation out of control and the job being killed.

The condition comes up when LHE3 weights are enabled and Pythia8 does not have any shower weights to provide. In that case, this loop begins at 1 and the stopping condition is that the iterator (being increased) is equal to zero, which won't happen for a very long time (the int has to flip over and reach zero again). This has caused a number of test failures already.