Fix tests on some platforms

[vbarbaresi]

• For Python 3.10: install a protobuf compiler for pycld3 build
  It turns out that pycld3 doesn't have prebuilt wheels yet for Python 3.10, so we need the tools to build it.
  I just saw that you already opened an issue for that Wheel building fails for Python nightly (3.10) bsolomon1124/pycld3#18

• Using % 256 == 0 in tests was actually masking failures: 256 return code meant that the binary wasn't found
  Explicitly call ./bin/trafilatura or ./Scripts/trafilatura so that we don't have rely on the binary being in the PATH.

  Also use the more modern subprocess.run() instead of os.system.
  We could collect stdout to test more things but it didn't work the same on all platforms so I didn't add it and use the return code only.

• A weird encoding issue happened on Windows from processes spawned by multiprocessing.Pool (and only in this case)
  I could reproduce using export PYTHONIOENCODING='cp1252' before running tests.
  Setting PYTHONIOENCODING='utf-8' during the test fixes the issue.
  Maybe we should set it in the application, if someone reports encoding issues on Windows. I didn't want to fix something that wasn't broken so I just fixed the tests.

[codecov-commenter]

Codecov Report

Merging #124 (25d502b) into master (776d706) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #124   +/-   ##
=======================================
  Coverage   94.63%   94.63%           
=======================================
  Files          19       19           
  Lines        2648     2648           
=======================================
  Hits         2506     2506           
  Misses        142      142           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 776d706...25d502b. Read the comment docs.

[sourcery-ai]

Sourcery Code Quality Report

✅  Merging this PR will increase code quality in the affected files by 0.12%.

Quality metrics	Before	After	Change
Complexity	1.59 ⭐	1.62 ⭐	0.03 👎
Method Length	198.33 😞	203.00 😞	4.67 👎
Working memory	7.73 🙂	7.74 🙂	0.01 👎
Quality	58.85% 🙂	58.97% 🙂	0.12% 👍

Other metrics	Before	After	Change
Lines	346	357	11

Changed files	Quality Before	Quality After	Quality Change
tests/cli_tests.py	58.85% 🙂	58.97% 🙂	0.12% 👍

Here are some functions in these files that still need a tune-up:

File	Function	Complexity	Length	Working Memory	Quality	Recommendation
tests/cli_tests.py	test_sysoutput	5 ⭐	288 ⛔	10 😞	48.67% 😞	Try splitting into smaller methods. Extract out complex expressions
tests/cli_tests.py	test_cli_pipeline	1 ⭐	662 ⛔	9 🙂	50.89% 🙂	Try splitting into smaller methods
tests/cli_tests.py	test_parser	2 ⭐	322 ⛔	6 ⭐	59.64% 🙂	Try splitting into smaller methods
tests/cli_tests.py	test_input_filtering	0 ⭐	236 ⛔	7 🙂	62.65% 🙂	Try splitting into smaller methods

Legend and Explanation

The emojis denote the absolute quality of the code:

• ⭐ excellent
• 🙂 good
• 😞 poor
• ⛔ very poor

The 👍 and 👎 indicate whether the quality has improved or gotten worse with this pull request.

---

Please see our documentation here for details on how these metrics are calculated.

We are actively working on this report - lots more documentation and extra metrics to come!

Help us improve this quality report!

[vbarbaresi]

oops I just realized that I removed the assert here @adbar
I initially wanted to assert on result but I changed my mind and didn't add the assert back

[adbar]

should be fixed in 07216a3

[adbar]

error on Windows for the line that wasn't part of the tests...

[vbarbaresi]

oh right, it's the same issue that I fixed in the multiprocess pool workers.. So adding PYTHONIOENCODING in the run command environment should fix the issue:

subprocess.run([...], env={"PYTHONIOENCODING": "utf-8"})

But I'll try on a Windows machine this week-end.
I want to see if it should be set globally in the app or if it's just a test issue on GitHub worker configuration.
I suspect we have to fix it globally and not just in the tests

[adbar]

ok!