Made it consistently read the PO file, decode its contents and then
check for the non-breaking space Unicode code point.
Previously we were erroneously skipping the interpretation of what we
read as UTF-8 text.
This was causing the test to fail on Windows with Python 3.5.