Avoid treating extensionless logs as CSV

[Shivansh Sharma (SHAI-shivansh-sharma)]

Summary

• Avoid treating extensionless plain-text logs from streams/stdin as CSV tables when the content has no CSV-style delimiter.
• Keep extensionless comma-delimited CSV converting to a Markdown table.
• Keep explicit .csv inputs on the existing CSV converter path.

Fixes #2474.

Test

• uv run --with pytest --with-editable . --python 3.12 pytest tests\test_csv_detection.py tests\test_csv_line_endings.py tests\test_csv_blank_runs.py

[Shivansh Sharma (SHAI-shivansh-sharma)]

@microsoft-github-policy-service agree

[Shivansh Sharma (SHAI-shivansh-sharma)]

Thanks for the detailed check. You were right that the original .sql / .log extension-list change did not materially improve the normal CLI or MarkItDown().convert(...) path.

I updated the PR to target the stronger issue you found instead: extensionless log/plain-text streams being inferred as CSV and rendered as a one-column Markdown table. The updated patch now downgrades content-only text/csv guesses to text/plain when there is no CSV-style delimiter evidence, while keeping extensionless comma-delimited CSV converting as a table.

I also added focused coverage for both cases and reran the existing CSV regression tests.