fix: align qformat and harden utf8 regression test

qformat now trims the shared prefix against both tag lines instead of
 checking first_tags twice, which keeps the generated ? lines aligned when the
 second tag line has different leading spaces.
The UTF-8 regression test now asserts ? lines and ^ markers so it
 exercises the intraline diff path instead of passing on a plain replace.
Rename the UTF-8 regression test to match the PR description.
This release stays on 0.5.1 because the UTF-8 intraline diff fix changes
 generated diff output for some inputs.
The public API is unchanged, but diff formatting behavior differs, so we
 treat this as a release-line bump.

00-007b6df9

Author: RandyMcMillan

tests/tests.rs

@@ -141,6 +141,8 @@ fn test_differ_compare_utf8() {
     let result = differ.compare(&first_text, &second_text).join("");
     assert!(result.contains("🦀"));
     assert!(result.contains("🦞"));
+    assert!(result.contains("? "));
+    assert!(result.contains("^"));
 }
 
 #[test]