Description

Adds ArabiziConverter, a deterministic PromptConverter that transliterates Arabic script into Arabizi (Latin-script "chat Arabic"), where letters with no Latin equivalent are written with shape-resembling digits (HAH -> 7, AIN -> 3, QAF -> 8). It applies a per-character mapping with Gulf-leaning conventions; no language model is involved, so the same input always produces the same output. Short-vowel diacritics and the tatweel connector are dropped, and non-Arabic text (Latin, digits, punctuation) is left unchanged. The mapping is intentionally lossy, mirroring how Arabizi is actually written.

The mapping follows the documented Arabic chat alphabet (Gulf-leaning where regional variants exist, e.g. QAF -> 8 with GHAIN -> gh to avoid the regional 8 collision). Feedback on specific letter choices is welcome.

Fourth in the set of atomic Arabic-script converters, following BidiConverter (#1832), TatweelConverter (#1869), and ArabicPresentationFormConverter (#1888). It can later migrate to a shared CharacterSubstitutionConverter base alongside UnicodeConfusableConverter.

cc @romanlutz

Tests and Documentation

• Added tests/unit/prompt_converter/test_arabizi_converter.py: word transliteration, number-letters, multi-character mappings, dropped diacritics/tatweel, non-Arabic passthrough, mixed text, empty input, determinism, and unsupported-input-type rejection. All pass: uv run pytest tests/unit/prompt_converter/test_arabizi_converter.py
• Registered in pyrit/prompt_converter/__init__.py (import + __all__).
• Added a usage example to doc/code/converters/1_text_to_text_converters.py and regenerated the paired .ipynb plus the converter modality table in 0_converters.ipynb via JupyText.
• ruff and ty are clean; the converter-documentation conformance test passes.