The "loggf cut" in levels_lines.py doesn't work as intended

[andrewfullard]

This will arise after #414 is merged. I added the ability to use the A_ul Einstein coefficient from the Chianti data directly in the case where gf is zero (and thus loggf = -inf). The inverse should also be true, retaining the original behaviour of Carsus i.e. that the lines with low loggf are cut. It is likely that this breaks for CMFGEN data.