fix: preserve CJK/Unicode characters in directory listing for non-browser clients

[zlvb]

Problem

When browsing directory listings from non-browser HTTP clients (e.g. Nintendo Switch DBI), CJK characters (Chinese, Japanese, Korean) in file and directory names are displayed as HTML entity codes like 巨丝事 instead of the actual characters.

This happens because he.encode() converts all non-ASCII characters into numeric HTML entities (e.g. 巨 → 巨). Standard web browsers decode these entities automatically, so the issue is invisible in browsers. However, lightweight HTTP clients that don't implement full HTML parsing display the raw entity codes, making CJK filenames completely unreadable.

Root Cause

1. he.encode() over-escapes non-ASCII characters: The he library's encode() function converts all non-ASCII characters (including CJK) to &#x...; HTML entities. Non-browser clients display these entities as-is.

2. Missing charset=utf-8 in Content-Type header: The directory listing response header was set to text/html without specifying charset, which can cause encoding issues in clients that don't sniff the <meta charset> tag.

3. ensureUriEncoded() function bug: An unreachable return statement on line 39 prevented the URL encoding logic from ever executing, causing non-ASCII characters in redirect URLs to remain unencoded.

Fix

• Replace he.encode() with a custom escapeHtml() function that only escapes the 5 HTML-unsafe characters (& < > " ') while preserving CJK/Unicode characters as raw UTF-8.
• Add charset=utf-8 to the Content-Type response header for directory listings.
• Remove the dead return text line in ensureUriEncoded() so URL encoding works correctly.

Files Changed

• lib/core/show-dir/index.js - escapeHtml() + charset fix
• lib/core/index.js - ensureUriEncoded() bug fix