fix: apply the download_file stub to moon_engine too (#160)

[XEDAB]

Description

Close #160
Resolved the issue in #160 by modifying the erroneous line 66 in conftest.py (monkeypatch.setattr(moon_cli, "download_file", fake_download))) (which could potentially cause the engine to attempt a real download) to monkeypatch.setattr(host, "download_file", fake_download)
and moved it into the for host in (moon_engine, moon_cli): loop, while adding a comment stating that this monkeypatch statement should not be moved.
This modification only involves conftest.py and does not involve any other changes.

Type of change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would change existing behavior)
• Documentation update
• Refactor / code cleanup
• Other:

Checklist

• I have tested my changes locally
• If this affects shared logic (extraction, download engine), I also
  applied the equivalent change to moon_cli.py
• I have kept the single-file architecture (no package split)
• I have not added new dependencies without justification in the PR
  description

Screenshots / logs (if applicable)

↑Normal state ↑The state after commenting out lines 181 and 255

[XEDAB]

By the way, I looked at line 270 mentioned in #160; it is a comment. I commented out the two lines of self._inc("ok") (181 and 255) in the fuckingfast branch, and the test crashed as expected.

[LeyckerS]

Correct, and verified rather than assumed:

ENGINE download_file: fake_download
CLI    download_file: fake_download

Both front-ends now hold the stub; before this, the engine held the real function. pytest tests/ -q → 50 passed.

Your verification in #160 is the part I want to point at. You did not just move the line — you went to moon_engine.py:270, found it was a comment rather than the counter, located the real increments at 181 and 255, annotated those instead, confirmed the tests went red, and restored. The line number in my issue was wrong and you worked past it without making a fuss. That is the third time you have had to correct something I wrote, and you have been right every time.

Putting the stub inside the loop rather than adding a second setattr beside it is also the right shape: the bug was that one line sat outside the loop, so the fix is to make the loop the only place any of this happens. Your comment says exactly that, which is what stops it recurring.

One cosmetic thing I am not sending you back for: that comment is a single 228-character line, against line-length = 100 in ruff.toml. It only passes because E501 sits in the ignore list, and every other comment in the file wraps. I will wrap it in the bookkeeping commit rather than cost you a round trip — the wording is yours and I am not changing it.

Merging.