fix: write StarDict metadata as bytes on Python 3 - #149

Problem

Fixes #91.

export_stardict() opens its .ifo file in binary mode but writes text strings to it. On Python 3 even a single ASCII entry raises TypeError: a bytes-like object is required, not 'str', leaving an incomplete export.

Change

Write the metadata with byte-string templates, preserving the existing binary file mode and LF output. Keep the book title encoded as UTF-8 and encode the generated date as ASCII. The word ordering and .idx/.dict paths are unchanged.

Add two standard-library regression tests covering an empty dictionary and a multilingual export. They check the UTF-8 title, case-sensitive sorting tie-break, binary offsets/lengths, actual index size, and metadata contents. No test dependency or dictionary download is needed.

Verification

python -B -m unittest discover -s tests -p test_stardict_export.py -v

• Before the fix: FAILED (errors=2), both at the first .ifo.write().
• After the fix: Ran 2 tests / OK on Python 3.12.14, Python 3.14.7, and Linux Python 2.7.18 (python:2.7-slim).
• Reverting only the production file in an independent copy restores the two original errors; the regression tests remain present.
• git diff --check passes.

The reproducer and tests above were executed locally.