Is it necessary to convert valid InverseVariance uncertainties to StdDevUncertainty?

In `09b-Specutils/Specutils_analysis.ipynb`, an example SDSS spectrum is read. SDSS stores uncertainty as `InverseVariance` by default, and the SDSS reader is aware of this. However, in the subsequent analysis examples, the uncertainty has to be converted to standard deviation.

* Is it still the case that the analysis functions can only accept `StdDevUncertainty`?
* If not, can we simplify the notebook by directly using `InverseVariance`?
* See also astropy/specutils#1298.