fix: strip invisible word joiner from OCR output

[pa4uslf]

Summary

• strip U+2060 WORD JOINER from OCR decoded output only
• keep the upstream dictionary unchanged to avoid label index shifts
• add a targeted regression test for CTC decode behavior

Testing

• PYTHONPATH=. pytest -q -o addopts="" tests/unittest/test_rec_postprocess.py

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[pa4uslf]

I have read the CLA Document and I hereby sign the CLA