fix(renderer): skip a body tag inside head scripts when measuring body text

[us]

ref #605

html_body_text_len found the opening tag with find("<body"), which matches the first mention anywhere in the document, including inside a <script> in <head>. A loader that writes <body><script src=x><\/script></body> into an iframe made the real body measure 0 characters, so a fully loaded page was classified as a thin render and escalated to the next tier.

• New find_body_open skips comments and the content of script, style, template and textarea, and steps over whole tags so a <script or <!-- inside a quoted attribute value is not read as a raw-text region.
• Falls back to the previous plain find("<body") when the scan finds nothing (unterminated comment or raw-text element).
• The closing-tag search and the counting loop are unchanged.
• Tests: the reproduction from the issue (0 before, 17 after), raw-text and comment skipping, case-insensitive tags, <bodyguard>, attribute-value lookalikes, empty comments, unterminated cases.

Checked against the page from the issue, rendered HTML of about 1.16 MB: the first <body sits inside the head script (the measured span was 49 characters, all tags), the scan now lands on the real body.

No config, API or response shape changes. Pages that were measured correctly before measure the same; only pages whose first <body mention is not the real tag change.