fix: handle utf-8 chunk boundary detection

[cnYui]

Summary

• Fixes (bug): Unicode characters might break chunk decoding logic leading to [Binary File] #578 by treating a UTF-8 sample chunk ending inside a multibyte character as text instead of binary.
• Keeps the change limited to the existing decode heuristic; full file reading and output formatting are unchanged.
• Adds a regression test for a 1024-byte boundary that splits a 3-byte UTF-8 character.

Test plan

• RED: python -m pytest tests/test_filesystem.py::test_content_keeps_utf8_text_when_multibyte_character_crosses_chunk_boundary -q
  • Before the fix: failed because node.content returned [Binary file].
• GREEN: python -m pytest tests/test_filesystem.py::test_content_keeps_utf8_text_when_multibyte_character_crosses_chunk_boundary -q
  • After the fix: 1 passed.
• python -m pytest tests/test_filesystem.py tests/test_ingestion.py -q
  • 9 passed.
• python -m ruff check src/gitingest/utils/file_utils.py tests/test_filesystem.py
  • All checks passed!.
• python -m ruff format --check src/gitingest/utils/file_utils.py tests/test_filesystem.py
  • 2 files already formatted.
• python -m pytest -q -k "not bitbucket"
  • 154 passed, 7 deselected.

Note: full python -m pytest -q currently fails on three Bitbucket-hosted tests/query_parser/test_git_host_agnostic.py cases because git ls-remote https://bitbucket.org/na-dna/llm-knowledge-share HEAD prompts for credentials in this non-interactive environment (fatal: could not read Username for 'https://bitbucket.org'). The failure is outside this file-content decoding path.