feat(text-normalizer): add Cyrillic and Greek profiles and multi-script fixtures

[mohaelmrabet]

Summary

Extends the profile set beyond Latin and Arabic, and adds a fixture file so a normalization sample can be added without touching PHP.

Cyrillic

NormalizerProfile::cyrillic() folds the letters no normalization form unifies: Yo (ё → е), the Ukrainian/Belarusian і, ї, ў and ґ. Combining marks are stripped by NFD as elsewhere.

Greek

NormalizerProfile::greek() folds final sigma (ς → σ) and strips tonos through canonical decomposition, so Αθήνα and ΑΘΗΝΑ fold together.

Fixtures

tests/fixtures/normalization_samples.json holds one original/expected pair per script, walked by testMultiScriptJsonFixtures. all() now composes the four profiles, so its name becomes arabic_search_latin_cyrillic_greek.

Test plan

• composer test — 51 tests, 101 assertions
• composer analyse — no errors

Note

The second commit fixes two Greek expectations that had ς surviving inside a word while folding when it stood alone — the character map folds a code point wherever it appears, and that is the behaviour the profile wants. It also types the fixture rows, which PHPStan rejected at level max (file_get_contents returns string|false, json_decode returns mixed).

Worth a look

cyrillic() maps і and ї to the Latin i, so Київ normalizes to киiв — a token mixing two scripts, and Киев (киев) stays distinct from Київ even though they are the same city. Folding to the Cyrillic и would collapse them instead. The tests and fixtures currently pin the Latin behaviour; happy to change it if the collision is what you want.