Add model selection to LLM provider configuration

[Halone228]

Describe feature

Problem

When configuring a custom LLM provider (e.g., OpenAI-compatible API via a custom baseURL), users
can set the API endpoint and key, but cannot choose which model to use. The system always falls back
to the provider's hard-coded default model.

This is problematic for:

• Self-hosted or proxy endpoints (Ollama, vLLM, LiteLLM) where the available model names differ
  from the provider's built-in defaults.
• Users who want to pin a specific model (e.g., GPT-4.1 vs GPT-4o) for a given provider
  configuration.
• Providers that expose multiple model families under the same API base.

Current behavior

1. Go to Options → AI / LLM → Add provider.
2. Select a provider type and enter a custom Base URL.
3. Save.
4. The chat always uses the provider's built-in default model, ignoring any custom base URL model
   offerings.

Expected behavior

The "Add provider" dialog should include a Model field (optional) that lets the user specify
which model ID to use by default for this provider configuration. If left blank, the current
behavior (fallback to provider default) should be preserved.

This is a small, backward-compatible change that significantly improves flexibility for custom
endpoints.

Additional Information

No response