[binhvq]

Fixes #354.

do_expand_tabs measured the Unicode width of everything between tabs, so control characters threw the padding off. It now tracks the column like GNU diff 3.12 does with -t:

• \r goes back to column 0. In the normal and context formats the line marker (< , ! , ...) is written again after it when more text follows; unified and ed don't repeat theirs, same as GNU.
• \b goes one column back. At column 0 it is dropped, which is also what GNU prints.
• other control characters, C1 included, take no column
• invalid UTF-8 bytes take one column each, instead of the whole chunk being counted by bytes when one byte is invalid

Text runs between control characters are still measured with UnicodeWidthStr, so the ZWJ emoji test keeps its width. do_write_line takes the marker as a new argument.

Checked against GNU diff in en_US.UTF-8 for the normal, unified, context and ed formats. In the C locale GNU counts every non-ASCII byte as one column; that was already different here and isn't touched.

[github-actions]

GNU diffutils testsuite comparison:

Test results comparison:
  Current:   TOTAL: 33 / PASSED: 20 / FAILED: 9 / SKIPPED: 4
  Reference: TOTAL: 33 / PASSED: 19 / FAILED: 10 / SKIPPED: 4

Changes from main branch:
  TOTAL: +0
  PASSED: +1
  FAILED: -1

Test improvements (1):
  + expand-tabs