Bug in noodlification, specifically in simulation reduction

[jurajsic]

There seems to be some problem with reduction of the intersection automaton during noodlification, see this commit for an example where this shows up.
The intersected automaton with epsilons should be simple line q0 -0-> q1 -0-> q2 -eps-> q3 -0-> q4 -0-> q5 but for some reason, after reduction (line 382), it turns into q0 -0-> q1 -0-> q2 -eps-> q3 with self-loop on q3 and NO final states.
Maybe the problem is that we are reducing and putting it into the same variable?

[Adda0]

I will have a look. Thank you for the bug report.