fix: set view blob charset=utf-8 to avoid non-ASCII mojibake

[wang1212]

Problem

When users click the View as Markdown action, the page opens in a new tab with non-ASCII text (e.g. CJK content) rendered as mojibake.

Root cause

viewAsMarkdown() opens the page content in a new tab via:

const blob = new Blob([contentToView], { type: "text/plain" });
const url = URL.createObjectURL(blob);
window.open(url, "_blank");

Blob encodes the JS string as UTF-8 bytes, but type: "text/plain" does not declare a charset. Per the HTTP/text/* default, the browser falls back to ISO-8859-1 (latin1) when decoding the resulting blob URL, so UTF-8 byte streams are decoded as single-byte latin1 characters — that's the mojibake.

Fix

Declare charset=utf-8 on the blob type:

-      const blob = new Blob([contentToView], { type: "text/plain" });
+      const blob = new Blob([contentToView], { type: "text/plain;charset=utf-8" });

The content is already UTF-8 (it's produced by the existing markdown extraction from the page DOM and from any generated per-page .md files), so declaring the charset only fixes decoding and changes nothing else.

Impact

view action only. copy and the AI-tool actions are unaffected (clipboard write and link construction already handle UTF-8 correctly).

Tested on

• A Docusaurus 3.10 site with i18n.defaultLocale: zh-Hans, Chinese documentation pages.
• Before: view tab shows 应用安全 etc.
• After: view tab shows 应用安全 correctly.

/cc maintainers — happy to adjust if you prefer a different approach (e.g. serve via a generated .md route when generateMarkdownRoutes is enabled).