Keep changed code blocks inside a fence

[tanghong123]

Fixes #62.

Two small changes in src/index.ts.

1. postprocess() dropped the <code> child of <pre>.

Assigning the unified diff to pre.textContent replaces every child of the <pre>, including its <code>. Pandoc's HTML reader only treats <pre> as a code block when it wraps a <code>, so the block round-tripped back as a paragraph — fence lost, lines reflowed, and smart punctuation rewriting the contents, so --config rendered as –config and "world." as “world.”. That last part is what makes this more than cosmetic: a command copied out of a diff can be silently wrong.

$ printf '<pre class="diff">-a --config\n+b --config</pre>' | pandoc -f html -t markdown
-a --config +b --config

$ printf '<pre class="diff"><code>-a --config\n+b --config</code></pre>' | pandoc -f html -t markdown
``` diff
-a --config
+b --config

**2. The intermediate conversion ran with syntax highlighting on.**

Pandoc then emits one `<span>` per code line and keeps the newlines as text nodes between them, which `node-htmldiff` discards — so a changed block carrying a language tag collapsed onto a single line even with the `<code>` restored. Plain blocks were unaffected, which is why `test/old.md` (an indented block) only ever exposed the first bug.

`--no-highlight` applies to the intermediate conversion only; the final render is a separate Pandoc call and still highlights, so ``` diff output stays colourised. I used `--no-highlight` rather than `--syntax-highlighting=none` for compatibility with older Pandoc, but 3.10 deprecates it and warns — happy to switch, or select by version, if you'd prefer.

**Verification**

`pandiff test/old.md test/new.md` on unmodified `master` does not reproduce the committed `test/diff.md`:

print(“Hello”) -print(“world.”) +print(“world!”) print(“Lorem ipsum
dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor
incididunt”)

With this patch it matches the golden file, apart from the image paths that the test harness supplies via its `extract-media` / `resource-path` options.

I could not use `npm test` as a gate: it reports 0 passing / 27 failing on unmodified `master` in my environment (ENOENT on `test/*.md`, timeouts), identically before and after this change, so those failures look environmental. Your CI may differ.

🤖 Generated with [Claude Code](https://claude.com/claude-code)