better implementation for `isinf`

[simeonschaub]

Thanks for this package! I find it really useful for our physics lab, since all the data we get is garbage anyways, so thanks to GarbageMath, I know right away that our result will be garbage too. No more annoying Gaussian error propagation and guessing numerical uncertainties until the error bars for the plot look nice! I have always found dealing with large numbers to be quite frustrating though; I think for all practical application treating 3 as infinity should be reasonable enough.