Detect HTML in maybe_is_html regardless of case or leading whitespace

[MohammadHijjawi97]

maybe_is_html compared the first 4 bytes against {b"<htm", b"<!DO", b"<xsl", b"<!X"}, so only an uppercase <!DOCTYPE or a lowercase <htm was recognized. These inputs were all missed:

• a lowercase <!doctype html>, the common HTML5 form;
• an uppercase <HTML> tag;
• a UTF-8 BOM;
• leading whitespace or newlines.

Docs.aadd_file wrote such files to a .txt temp file and indexed the raw markup instead of parsing it with html2text.

>>> maybe_is_html(BytesIO(b"<!doctype html><html></html>"))
False

This change reads a small prefix, drops a UTF-8 BOM and leading whitespace, and compares case-insensitively.

Added a parametrized offline test. It covers lowercase and uppercase doctypes and tags, leading whitespace and a BOM, plus PDF and plain-text negatives, and checks that the file is rewound. The four new positive cases fail on main and pass here. ruff and black are clean.