Do not try to call uname on Windows during versioninfo(true)

[tkelman]

Add a test that versioninfo(IOBuffer(), true) works

Yay, posix assumptions. mprsv also brings up memories of a past life where I spent too many months dealing with a manifold pressure regulating shutoff valve...

[vtjnash]

it would be better if you could assert some of the content is correct, but i guess this test is better than none. i think uname probably was previously getting provided by git?

[tkelman]

Agreed, and correct. Any particular content we care about enough to test often? I just couldn't be bothered picking. Though it looks like the Pkg.status part of this will go and clone METADATA so we can't actually put this in test/misc.jl.

[vtjnash]

Pkg.status probably should never be a mutating operation, but perhaps that'll be a Pkg3 improvement.

I would usually just look for a bit of the structure -- something at the beginning and something at the end -- to check that it wasn't a complete no-op failure.

[tkelman]

Yeah. Sounds like a good idea. For now I'll move it to test/pkg.jl because why not. I also think there might be some other copy of uname floating around on the appveyor path so this particular issue would only be noticed locally.