Issue with Unicode characters in the test description

With some setups when the test description (after the $) contains Unicode characters (such as accents), should.py fails with the following error:

Traceback (most recent call last):
  File "should.py", line 1151, in <module>
    status = fs.test(variables = variables, cd = args.cd, cd_same = args.cd_same, output = args.output, verbose = args.verbose, only = only)
  File "should.py", line 1024, in test
    s.test(open(f), variables, verbose - 1, only=only)
  File "should.py", line 735, in test
    for l in list(should_lines) + [DIRECTIVE_SCRIPT]:
  File "/usr/lib/python3.5/encodings/ascii.py", line 26, in decode
    return codecs.ascii_decode(input, self.errors)[0]
UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 in position 16: ordinal not in range(128)

For instance it fails with Debian docker images (Debian 7, 8 or 9, but not with version 10).

Here is a Dockerfile to reproduce the bug:

from debian:9-slim

run apt-get update && apt-get install -y python3 wget
run wget -O should.py https://gitlab.bioinfo-diag.fr/vidjil/should/raw/master/src/should.py?inline=false

run echo 'ls should.py\n\n$ ÉÀ\n1:should.py' > test.should

And then

# Build the image
docker build -t should/issue:47 - < Dockerfile
# Launch the test
docker run should/issue:47 python3 should.py test.should

This script is used with CI which can rely on such Debian images.