feat: add Ollama Cloud provider support

[saurondark22]

This PR significantly upgrades the Ollama integration by adding a new cloud provider option, and disabled thinking feature for reasoning-capable models.

Changes:

Ollama Cloud Provider (new)

• Added OllamaCloudProvider class API key entry just similar to gemini.
• Provider-specific after_load() hook handles endpoint selection and auth configuration
• pre-added "gemma4:31b", "gemma3:12b", "deepseek-v4-flash", "nemotron-3-nano:30b" and "gpt-oss:20b" models to select, also option for custom models.

Local Ollama Improvements

• Exposed three user-configurable settings: num_ctx (context window), num_predict (max tokens), temperature
• Disabled the think step on reasoning-capable models to eliminate wasted latency on internal chain-of-thought when not needed
• Added special handling for gpt-oss since Ollama only accepts string think levels — maps "low" to approximate "off"

Files modified:

• aiprovider.py — major expansion with cloud provider and local improvements
• WritingToolApp.py — registration and wiring for cloud provider
• ui/SettingsWindow.py — new UI controls for Ollama settings
• ui/OnboardingWindow.py — cloud provider in onboarding flow
• locales/en/LC_MESSAGES/messages.po — new translation strings

Why this matters:

• Cloud support: Users can now use managed Ollama endpoints (e.g., Ollama Cloud), no need to install Ollama or manually manage models.
• Performance control: Power users can tune context windows and token limits instead of fighting hidden defaults
• Reasoning model UX: Disabled "thinking" for ollama local and cloud for better latency when users just want a quick rewrite/summary