Preserve raw filename bytes in trashinfo Path entries

[SOV710]

Summary

Fix .trashinfo Path= handling for filenames that are valid filesystem byte sequences but not valid UTF-8.

trash-put can now trash files with names such as b'\xa4-\xc8\xcf-\xc1+\xb8.txt' without losing the original filename bytes.

Bug

trash-put previously percent-encoded Path= from a Python text string. On Python 3, non-UTF-8 filenames are represented with surrogate escapes, and trying to encode those surrogates as UTF-8 can raise UnicodeEncodeError while generating the .trashinfo file.

This means a file can exist on disk and be passable to os.*, but still fail during trash-put.

Reproduction

python3 - <<'PY'
import os

name = b'\xa4-\xc8\xcf-\xc1+\xb8.txt'
fd = os.open(name, os.O_WRONLY | os.O_CREAT | os.O_EXCL, 0o600)
os.close(fd)
print(os.fsdecode(name))
PY

Then pass the printed filename to:

trash-put ''

Before this PR, trash-put could fail while writing the .trashinfo metadata.

Fix

This PR makes .trashinfo path encoding byte-preserving:

• Write Path= with os.fsencode(path) followed by byte-based percent encoding.
• Read Path= with byte-based percent decoding followed by os.fsdecode(...).
• Use the same helper in both parse_path() and ParseTrashInfo so list/rm/restore parse paths consistently.

This keeps .trashinfo contents UTF-8-compatible while preserving the original filesystem bytes.

Tests

Added coverage for:

• Formatting non-UTF-8 path bytes into percent-encoded Path=.
• Parsing percent-encoded Path= back to the original filesystem bytes.
• ParseTrashInfo callback behavior for non-UTF-8 paths.
• An e2e trash-put regression test with a real non-UTF-8 filename.

Verified with:

pytest tests/test_trashcli_lib/test_parsing_trashinfo_contents.py
pytest tests/test_put/cmd/e2e
uv run --no-project --with pytest --with flexmock --with six --with psutil --with typing_extensions python -m pytest tests/test_put/cmd/test_put.py