Error in collisional Redl-Sauter conductivity when Zeff < 1

[hoppe93]

Under some circumstances, it seems that Zeff may be evaluated to be < 1 by just a little little bit (probably due to numerical problems), which causes problems for the Redl-Sauter conductivity when collisional effects are included. The problems are due to that sqrt(Zeff-1) is evaluated explicitly in the code, which will result in a NaN (or floating-point error in debug mode).

Since we don't expect Zeff < 1, we should just be able to set Zeff = 1 if it's less than 1 prior to the conductivity computation.