fix(security): fix percent_decode to handle multi-byte UTF-8 properly

The previous implementation decoded percent-encoded bytes one at a time
and cast each to char, which produces invalid characters for multi-byte
UTF-8 sequences (e.g. Chinese filenames). Fixed to accumulate raw bytes
first and convert to String at the end with proper UTF-8 validation.

src/plugins/http/routes.rs

@@ -1624,29 +1624,29 @@ fn file_mkdir_route(
 }
 
 fn percent_decode(s: &str) -> String {
-    let mut result = String::with_capacity(s.len());
+    let mut bytes = Vec::with_capacity(s.len());
     let mut chars = s.bytes();
     while let Some(b) = chars.next() {
         if b == b'%' {
             let h = chars.next().unwrap_or(0);
             let l = chars.next().unwrap_or(0);
             let hex = [h, l];
-            if let Ok(s) = std::str::from_utf8(&hex) {
-                if let Ok(val) = u8::from_str_radix(s, 16) {
-                    result.push(val as char);
+            if let Ok(hex_str) = std::str::from_utf8(&hex) {
+                if let Ok(val) = u8::from_str_radix(hex_str, 16) {
+                    bytes.push(val);
                     continue;
                 }
             }
-            result.push('%');
-            result.push(h as char);
-            result.push(l as char);
+            bytes.push(b'%');
+            bytes.push(h);
+            bytes.push(l);
         } else if b == b'+' {
-            result.push(' ');
+            bytes.push(b' ');
         } else {
-            result.push(b as char);
+            bytes.push(b);
         }
     }
-    result
+    String::from_utf8(bytes).unwrap_or_else(|e| String::from_utf8_lossy(e.as_bytes()).into_owned())
 }
 
 fn list_video_files(dir: &Path) -> Vec<VideoFileInfo> {