fix: report accurate UTF-8 hook line numbers

[sylvesterkaczmarek]

Fixes #1251.

Read files as binary lines and decode each line independently so invalid UTF-8 is reported against the correct physical line. The hook now also continues after a decode failure and reports multiple invalid lines in one run.

Adds regression coverage for invalid bytes on multiple lines and for valid UTF-8 input, and records the fix in the changelog.

Validation:

• pytest tests/unit/test_require_utf8.py -q: 2 passed
• pre-commit checks on all changed files: passed
• git diff --check: passed