[refactor] increase consistency in stdout and stderr optional arguments

[wyattscarpenter]

I still think there's a good change to be had in here, but I'm trying a different approach. So, it stays as a draft PR for now.

On another note, sys.stdout and sys.strerr can themselves technically be None, and typeshed gives them MaybeNone to pay homage to this — I would like to actually handle this correctly in the mypy code here, but, what are you going to do, write a message to stderr about it? Guess we'll just roll on.

[github-actions]

According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅

[wyattscarpenter]

original comment, with the approach " [refactor] simply use default arguments for stdout and stderr"

Unless this runs me into the famous Python default argument folly, which doesn't seem likely to me, then this should be a better way to do it. (What, do we change sys.stdout during the course of the program? Guess CI will let me find out...) I guess you could argue that even trying to access a member variable like this sets us up for the footgun in the future...

Edit: "you could argue" — I've been convinced by my own argument. Also, I guess some of our test code, like a tiny amount, one test it seems, does actually rely on retargetting stderr or something.