GeezerVison

Summary

• Decode short MessagePack strings as UTF-8 when the JVM stores String values in compact byte arrays.
• Add a Turkish-character round-trip regression test.

Evidence

• Reproduced issue Non english characters not deserialized successfully #908 against the current source on Java 21 with -Dfile.encoding=windows-1252; the Turkish string decoded as mojibake.
• After the fix, MessagePackDataformatForPojoTest.testUnicodeStringRoundTrip passed in a credential-free, network-disabled Java 21 Docker container using the same charset setting.
• git diff --check passed. The project patch policy accepted 2 text files, 12 changed lines, and no binary changes.

Merge Danger

The change only selects UTF-8 for the short-string path, matching the existing UTF-8 decoder used for longer strings and compact-String fast paths.