[vbuch]

Fixes #267.

Japanese and other Unicode signature metadata was written with PDFString.of(), which loses characters when serialized. Encode non-ASCII Reason, ContactInfo, Name, and Location values with PDFHexString.fromText() (UTF-16BE with a BOM), while retaining literal strings for ASCII and empty values.

Uses the conditional encoding approach from #292 in a focused change based on current develop. Adds 14 save/reload regressions covering Japanese, accented Latin, Cyrillic, emoji, mixed text, ASCII, and empty strings with object streams both enabled and disabled. Includes rebuilt distribution files.

Validation:

• All 171 repository tests and repository-wide lint pass.
• All 29 PDF-LIB tests pass with 100% coverage.
• Built-package P12 signing smoke test preserves Japanese and emoji metadata and produces a signature whose byte range covers the document.
• Babel build, declaration generation with --skipLibCheck, and tsd pass.
• Plain tsc reports errors in installed @types/node declarations (including missing IteratorObject and AsyncIteratorObject); dependency declarations are unchanged by this PR.
• git diff --check passes.

[coveralls]

coverage: 100.0%. remained the same — codex/fix-unicode-signature-metadata into develop