Make the critical treshold user-definable instead of assuming double the treshold

[martinseener]

At moment the treshold given by the user is the actual warning value whereas the critical value is simply double the value (double standard deviation). Instead of assuming this make the critical value user-definable by optionally checking a fourth input parameter (if not given, fallback to double deviation behaviour)