Fix Models local register 500 INTERNAL_ERROR on Windows

[XYAIStudio]

Summary

Registering any local model from Models → LocalHardwarePanel returned a generic INTERNAL_ERROR 500. That was a shared register-path bug, not a bad GGUF.

Root causes:

1. ollama create Modelfile wrote FROM C:\Users\...\file.gguf without quotes. Backslashes/spaces break Ollama on Windows, so every GGUF import failed.
2. CLI output used text=True without encoding="utf-8" / errors="replace". On Chinese Windows (GBK) a failed ollama create raised UnicodeDecodeError, which is not OSError, so the API swallowed it as INTERNAL_ERROR.
3. OllamaModelManager.list_models() called .get() on the modern SDK ListResponse object. Probe already caught that (UI still showed “Ollama 已启用”); register did not, so a successful create still 500’d.
4. Already-pulled Ollama tags could not be registered (source=ollama was treated as “manual”, empty path failed validation, UI hid the button).

Fixes:

• Quoted POSIX FROM line + cwd = weight parent + UTF-8 CLI decode.
• Normalize ollama.list() dict and pydantic payloads.
• Register already-listed Ollama tags into the provider without ollama create.
• Recoverable failures return {ok: false, next_step} or LOCAL_MODEL_REGISTER_FAILED (400) with the real reason. Provider reload after a successful save no longer fails the request.

Target branch

• Base is develop (feature / fix — default)
• Base is main (requested: bugfix from latest main; do not merge / do not touch Release)

Type of change

• Bug fix
• New feature
• Breaking change
• Documentation
• Refactor / chore
• Release / hotfix

Test plan

• Added/updated tests
• uv run pytest tests/unit/test_local_register.py tests/unit/test_ollama_create.py tests/unit/api/test_local_models_register.py tests/unit/i18n/test_errors.py
• make lint + make typecheck
• make all (full suite not run in this environment)

Windows desktop verification (Ollama running)

1. Confirm the tray / http://127.0.0.1:11434 is up. Models page should show Ollama 已启用.
2. Already-pulled tag: an installed Ollama model should show 注册. Click it — it should succeed and become selectable in chat. No ollama create needed.
3. GGUF: search/pick a file such as C:\Users\<you>\Downloads\*.gguf (spaces in the path are fine) → 注册. Should import via ollama create and then appear on the local provider.
4. Recoverable failures (missing file, ollama create stderr, permissions) must show the real reason, not 服务器内部错误。

Checklist

• Updated CHANGELOG.md (if user-facing)
• README / docs updated (if needed)

Do not merge this PR and do not replace or cut a Release.