Epic 2 — Test nets (Hafs characterization + Warsh red cases)

[Hetchy]

Parent: #27 · Plan

Goal: Build the safety net before refactoring, and the Warsh target before implementing. Two distinct nets.

2A — Hafs characterization net (start now, parallel)

• Golden tests at the stable public boundary only: phonemes_str, letter_phoneme_mappings, char_phoneme_mappings, tajweed_mappings, phonetic_text — broad representative reference set (stops, muqattaat, lafdh al-jalalah, hamza wasl, tanween-at-waqf, idgham/ikhfaa/iqlab).
• Lock outputs so Epic 3a/3b can prove byte-identical Hafs behavior.
• Do not test internal classes slated for restructure (test behavior, not soon-to-change structure).

2B — Warsh expected cases (after 1b; red until Epic 4)

• Hand-author expected phonemes for 1–2 cases of every tajweed rule and every contextual pronunciation, prioritizing the Hafs deltas from 1b (things Hafs doesn't even test today).
• First-class tests for tajweed rules and mappings under Warsh.
• Keep red (xfail/skip) until Epic 4 turns them green.

Acceptance

• Hafs characterization net green, wired into CI as the refactor gate.
• Warsh expected-case suite committed, red, enumerating target behavior.