fix: render html in pull request comment bodies as markdown

[leynier]

Pull request comment bodies often contain a GitHub HTML subset (bot footers with <picture>, <img>, <a>, <sub>, <br>, HTML comments), which GptMarkdown does not render and showed up as literal text.

What changed:

• Add sanitizePullRequestCommentBody in lib/src/features/pull_requests/domain/pull_request_comment_body.dart to convert that HTML subset to Markdown before rendering, and wire it into PullRequestCommentMarkdown.
• Mirror the same sanitizer in mobile/lib/src/features/workbench/domain/pull_request_comment_body.dart and wire it into pull_request_conversation_section.dart.
• Handle images, dark-mode <picture> sources, links, inline formatting, blocks, breaks, <pre>, entities, and HTML comments while leaving fenced code blocks and inline code spans untouched.
• Cover desktop behavior with unit and widget tests, plus a mobile drift test for the shared footer and code-fence cases.

[pullfrog]

Important

HTML <blockquote> in a comment body throws while sanitizing, so that comment fails to render.

Reviewed changes

This pass reviewed the HTML-to-markdown sanitizer added for pull request comment bodies on desktop and mobile, plus the tests that pin bot-footer rendering.

• Comment HTML sanitizer — Converts the GitHub HTML subset (<picture>, <img>, <a>, formatting, blocks, entities) to markdown before GptMarkdown, leaving fenced and inline code untouched.
• Desktop and mobile wiring — PullRequestCommentMarkdown and the mobile conversation _CommentBody now sanitize immediately before render; the mobile file is the expected duplicate because alera_mobile cannot import the root package.
• Tests — Desktop unit tests pin exact conversions for the footer, images, dark <picture>, javascript: hrefs, entities, and fences; a widget test pumps a bot footer; mobile covers the footer and fence cases against drift.

  ^{｜ Fix all ➔ ｜ Fix 👍s ➔ ｜ View workflow run ｜ Using Grok ｜ 𝕏}

[pullfrog]

✅ No new issues found.

Reviewed changes

This pass reviewed the follow-up commit that fixes the blockquote sanitizer crash and tightens CRLF / <br> handling, plus the tests that pin those arms.

• Blockquote capture group — Both desktop and mobile converters now read group 1, matching the one-capture regex, so HTML <blockquote> no longer throws in build().
• CRLF and breaks — Prose segments drop a trailing \r, and <br> consumes one following newline so a hard break does not become an extra paragraph.
• Inline picture pass — _sanitizeInline no longer re-walks <picture>; the block pass already converted complete pictures, and the empty-src fallback still uses the inner <img>.
• Coverage — Unit tests now pin blockquotes (including empty), <pre>, strikethrough, inline <code>, headings, summaries, lists, Windows line endings, and pictures without a usable src.

  ^{｜ View workflow run ｜ Using Grok ｜ 𝕏}