Fix #134: incorrect optimization for qiskit circuit

[bichselb]

This should fix #134.

[dlyongemallo]

I noticed that this change has a side-effect on an edge case. Suppose the input is this:

   ┌───┐
q: ┤ T ├
   └───┘

Then, after applying optimisation, the previous output would've been no change:

   ┌───┐
q: ┤ T ├
   └───┘

However, with this change, the output now becomes:

   ┌─────────┐
q: ┤ Rz(π/4) ├
   └─────────┘

It's logically the same, but it's a change nevertheless. Is this desirable?

[jvdwetering]

I don't think that should lead to any problems. It is semantically still the same circuit.

[dlyongemallo]

I agree. I just thought I'd note it in case someone wanted a noop for an already optimised input, or if some tool which takes in qasm input treats t and rz(0.25*pi) differently for some reason.