Fix encoding error with print statements in doctests

[jwodder]

This fixes #3583. This fix was suggested by Stack Overflow user phd in
https://stackoverflow.com/a/50863820/744178.

[jwodder]

I would have added the minimal example from #3583 as a test case, but I wasn't sure how exactly to do that. Someone who knows what they're doing should probably do that.

[nicoddemus]

Hi @jwodder,

Thanks a lot for the PR, we appreciate it.

Could you please include a test which reproduces the problem you describe in #3583?

I suggest to take a look at the other tests in test_doctest.py for inspiration.

[nicoddemus]

This should be rst:

Fix encoding error with ``print`` statements in doctests.

[jwodder]

It is RST; reStructuredText uses single backticks for individual identifiers, while double backticks are for arbitrary inline code.

[nicoddemus]

Oh you are right, sorry;

We use inline code in the changelog though because there won't be any references to it in the final document (but TBH I don't remember if they render differently).

[coveralls]

Coverage increased (+0.05%) to 92.708% when pulling d382f3e on jwodder:fix-3583 into 4d0297b on pytest-dev:master.

[jwodder]

@nicoddemus I added a failing test case, but now the fix doesn't fix it; it was working when I tried modifying a copy of pytest in a virtualenv a few hours ago....

[jwodder]

I figured out what was wrong with the test case I added (the string should have been raw); it now correctly fails before applying the fix and succeeds after.

[nicoddemus]

LGTM, thanks again @jwodder!