Fix existing XML codec test to use Java-encoded strings

The test fed plain text, but .visustruct/.strk store text as
semicolon-separated code points, so it decoded to empty strings.
It had never run before CI existed.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Cp4EQgFQyGbM8MVLiWbJKQ

Author: claude

VisuStruct-iOS/VisuStructCore/Tests/VisuStructCoreTests/VisuStructXMLCodecTests.swift

@@ -2,12 +2,13 @@ import XCTest
 import VisuStructCore
 
 final class VisuStructXMLCodecTests: XCTestCase {
+    /// Texte stehen wie im Java-Editor als `;`-getrennte Unicode-Codepoints im XML.
     func testDecodesLegacyStruktogrammeditorXML() throws {
         let xml = """
         <?xml version="1.0" encoding="UTF-8"?>
-        <struktogramm caption="Beispiel">
+        <struktogramm caption="66;101;105;115;112;105;101;108;">
           <strelem typ="0" zx="0" zy="0">
-            <text>x = 1</text>
+            <text>120;32;61;32;49;</text>
           </strelem>
         </struktogramm>
         """