fix(fetch): suppress duplicated extracted comment text - #1863

Summary

• switch the HTML extraction path in FetchURL to inspect Trafilatura's main text and comments separately
• suppress extracted comments when they normalize to the same content as the main body
• add a regression test for duplicated GitHub issue extraction output

Root cause

trafilatura with include_comments=True can misclassify GitHub issue body content as comments and append a second malformed copy of the same text.

For pages like https://github.com/MoonshotAI/Moonlight/issues/4, this caused FetchURL to return the issue body twice:

• once as the normal extracted main content
• once again as a malformed "comment" block with missing spacing around inline code

Testing

• uv run pytest tests/tools/test_fetch_url.py -q
• uv run pytest tests/tools/test_tool_descriptions.py -q
• uv run pytest tests/tools/test_tool_schemas.py -q

Related issue

• Closes FetchURL duplicates GitHub issue body when comments extraction misfires #1862

Notes

• make test still reports an unrelated existing failure in tests/e2e/test_shell_modal_e2e.py::test_approval_consecutive_across_turns
• the fetch_url regression covered by this change passes