[jhdark]

Follow-up to #17 (#4).

IonisationChamber computed the concentration as flux_in / flow_rate if flow_rate > 0 else 0.0. Because nan > 0 is False, a NaN flow rate gave a signal of exactly 0, so a broken upstream block looked like a clean "nothing detected". It now gives a NaN signal; the flux still passes through unchanged.

ic = IonisationChamber()
ic.inputs[0], ic.inputs[1] = 100.0, float("nan")
ic.update(None)
ic.outputs[1]   # before: 0.0, after: nan

Zero and negative flow rates still give a zero signal, as before. I didn't make a negative flow raise: implicit solvers can try negative values while iterating, so raising inside update could stop a valid run.

Adds a regression test, which fails on the current code.

Open question, not changed here: detection_efficiency (and the value returned by func_efficiency) is only checked to be non-negative, so values above 1 are accepted. That's fine if it is meant as a calibration factor, but if it is strictly an efficiency, it should be limited to [0, 1].