Remove large-scale flow U(t) from BarotropicQG

[navidcy]

I'm tempted to remove the large-scale flow from BarotropicQG as it's too specific and makes the code hard to penetrate for first-time users. Perhaps move U(t) functionality in an example or a different module...

[glwagner]

alternatively you can make sol a NamedTuple and optionally activate U(t) evolution if sol.U isn't nothing

[navidcy]

@glwagner, I feel that FourierFlows/FourierFlows.jl#212 should be dealt first. Then what you suggest would be smoother.