Stop misdetecting Markdown that mentions HTML

[trsdn]

Fixes #14.

Independent PR — branches from master, touches only ClipboardContentDetector.cs.

Problem

GetMarkdownScore vetoed the entire score on any HTML-looking tag:

if (ContainsHTML(trimmed))
    return 0;

This only bites when the clipboard carries both text and HTML — which is exactly what editors and browsers publish. So Markdown that merely mentions a tag was classified as rich text, and Ctrl+Enter ran HTML → Markdown on a document that was already Markdown.

Change

• Markup inside fenced blocks and inline code is ignored, since angle brackets there are content, not markup.
• Remaining raw HTML is a penalty (-6) weighed against the other signals, not an absolute veto. Markdown that embeds a little raw HTML still wins on structure; text that genuinely is HTML still scores low and stays rich text.

Verification

All cases below have the same syntax-highlighted editor HTML on the clipboard alongside the text, which is what triggers the bug.

Clipboard text	Before	After
# Title + list + **bold**	Markdown	Markdown
# Layout + list + `<div class="card">`	RichText ❌	Markdown ✅
# Title + list + fenced ```html block	RichText ❌	Markdown ✅
<h1>Hi</h1><p>there</p> (HTML source as text)	RichText	RichText
Word paste: plain prose text + semantic HTML	RichText	RichText
HTML only, no text	RichText	RichText
empty	Unknown	Unknown

No regressions: the two cases that must stay RichText still do, including a paste whose plain-text side has no Markdown syntax at all.

dotnet build -c Release -r win-x64 succeeds with 0 warnings.

[trsdn]

Superseded by #53. These commits are already contained in that branch, so this PR would merge nothing on its own. Collapsing the stacked converter PRs into one avoids the base-rewriting problem that made the stack unmergeable.