Fix integrate.fixed_quad docstring to indicate None return value

[synapticarbors]

I noticed that scipy.integrate.fixed_quad return a tuple with the second element always equal to None, that is not documented (I had to look at the source code to realize why I was getting a tuple back instead of just a float). This has been in the code for a long time and it looks like previous versions of the docstring properly accounted for this, but the current version does not.

I assume there is some legacy reason for always returning a None, and a quick search of Github shows that most people use it as something like:

val = scipy.integrate.fixed_quad(...)[0]

I'm not sure my proposed docstring is optimally worded, but I'm happy to adjust it if others have suggestions.

[ev-br]

Not sure what the best wording is, but yeah, "a bit strange" is better than plain misleading. And this stray None is a bit strange. Merged, thank you @synapticarbors.

[synapticarbors]

Thanks @ev-br for merging my PR (this is my first contribution to scipy). I just assumed that since the None has been there for so long (since scipy 0.2 at least) that there must have been some interface that once existed with two useful return values and it's the way it is to not break code that assumed a tuple was returned.

[ev-br]

As a guess, the second return value might be there to parallel quad which returns the tuple (value, error estimate).