Extract every patch segment and decode byte lines - #78

Summary

extractcode --all-formats libmediainfo-0.7.43.diff raised sequence item 0: expected str instance, bytes found. patch_info() yields whatever the patch library returns, and that library returns bytes lines, while extract() joins them as text. The same loop also returned after the first patch segment, so a multi-patch file only produced one extracted file.

Changes

• Decode bytes lines from patch_text() with UTF-8 and fall back to Latin-1, then strip, so patch_info() always yields text lines.
• Move return [] outside the loop in extract() so every patch segment is written.
• Add two regression tests: a multi-segment patch file and the libmediainfo patch from the report.

Testing

• pytest tests/test_patch.py: 199 passed, 2 xfailed.
• Full suite: 77 failed, 441 passed, 33 skipped, 10 xfailed.
• On unmodified main in the same environment: 83 failed, 433 passed, 33 skipped, 10 xfailed. The six extra fixes are the patch tests that were failing because of the same byte decoding. The remaining failures need the 7z and rar extraction plugins, which are not installed here.
• ruff check and ruff format --check pass.
• Manual check: dnsmasq_2.63-1.diff now writes 19 subfiles for 19 segments, and unionfs-2.5.1_for_2.6.23.17.diff writes 43 for 43.

Closes #57