Convert byte column offsets to characters

[dpc00]

Fixes #25.

cppcheck reports the column ({column}) as a byte offset into the UTF-8 line, so every non-ASCII character before the error on the same line moved the highlight to the right, past the end of the line for errors near its end. Sublime counts characters.

reposition_match now converts the byte offset to a character offset using the source line (vv.select_line(line)), like the fix in SublimeLinter-xmllint. Both linters (cppcheck and cppcheck++) get it from a small abstract base class. ASCII lines are unchanged.

Checked with cppcheck 2.21.0 in Sublime Text 4215 (this branch loaded as SublimeLinter-cppcheck), on the file from the issue: with an e with acute accent and an emoji before them, the array error is now on the [ (start 18, it was 22) and the unread-variable error on the = (start 23, it was 27, i.e. on the line break).

Tests: tests/test_columns.py uses the real output of cppcheck for that file (source and output are in the test), for both linter classes, plus the converter on its own (multi-byte characters, an offset inside a character, past the end); 4 tests pass in Sublime Text 4215, flake8 with the repo config is clean.

🤖 Generated with Claude Code

https://claude.ai/code/session_01WtRdpWmSGFACg8HR3Vpk3d

[kaste]

Same as tslint, regression test should go through process_match

E.g.

   class TestRegex(unittest.TestCase):
       def assertMatch(self, string, expected, source):
           linter = Linter(sublime.View(0), {})
           match = next(linter.find_errors(string))

           # Use the supplied source as the main buffer, not an external file.
           match['filename'] = None
           actual = linter.process_match(match, VirtualView(source))
           self.assertIsNotNone(actual)

           # Ignore information we don't want to write down in the examples.
           self.assertEqual({k: actual[k] for k in expected}, expected)