[S2] Verify parser/turns_rounds.py _WORD_RE handles ASCII apostrophes; add regression test

[erinepshovel-code]

Tests/test_backend.py::test_tokenize_ascii_apostrophe_contraction verifies that stats.py::_WORD_RE tokenizes "don't can't it's" as whole tokens. The test was added to fix issue #37.

backend/src/edcmbone/parser/turns_rounds.py::_WORD_RE is a visually similar regex but has no equivalent regression test. It is possible that the same apostrophe fix was not applied to the parser regex, or that it was applied but is untested and could regress.

Action required:

1. Byte-compare stats.py::_WORD_RE and parser/turns_rounds.py::_WORD_RE to confirm whether the character class ['‘’] (or equivalent) is present in both.
2. If the parser regex is missing the apostrophe character class, apply the same fix that was applied to stats.py for issue stats.py _WORD_RE misses ASCII apostrophe — contractions split on raw input #37.
3. Add a test to Tests/test_backend.py (or a new file) that feeds "don't can't it's" through the parser tokenization path (not just stats.tokenize) and asserts whole-token output.

Without step 3, even if the regex is currently correct, the fix is unguarded and will not be caught if it regresses.