Release 1.5.0: Arabic, Japanese, and Turkish PII MLX support

[maziyarpanahi]

Summary

Release prep for OpenMed 1.5.0, focused on Arabic, Japanese, and Turkish PII support plus the completed MLX rollout.

What changed

• Adds SDK support for ar, ja, and tr PII extraction/de-identification defaults, language metadata, regex patterns, fake replacement data, and anonymizer locale routing.
• Registers all API-visible source PII checkpoints for the new languages: 2 Arabic, 3 Japanese, and 32 Turkish models, bringing the SDK PII registry to 247 models.
• Adds MLX preconverted routing for the 28 uploaded and collection-verified Arabic/Japanese/Turkish -mlx repos.
• Leaves the 9 unsupported ModernBERT, Qwen3, and Longformer PII checkpoints deferred until converter architecture support lands.
• Fixes Turkish street-address regex coverage for common name-first forms such as Atatürk Caddesi 12.
• Bumps package/docs/demo version surfaces to 1.5.0 and updates the changelog, README, docs, website copy, and multilingual smoke example.

Hub rollout status

• Converted/uploaded: 28 MLX repos
• Added to OpenMed/medical-mlx-models: 28
• Verified missing from collection: 0
• Deferred: 9 unsupported architecture repos
• Failed: 0

Verification

• uv run pytest tests/unit/test_pii_i18n.py tests/unit/test_model_registry_multilingual.py tests/unit/mlx/test_mlx_inference.py — 211 passed
• uv run pytest tests/unit/test_pii_i18n.py — 150 passed after the Turkish address fix
• uv run python scripts/release/check_release_version.py --version 1.5.0 — passed
• git diff --check — passed