I have a DataFrame full of floats (positive and negative) and some NaN. I'd like to replace every single float number with its sign: if it's NaN -> it remains Nan if positive -> replace with...