[Theia AI] Add LLMSettings to UI

[eneufeld]

Feature Description:

AI agents need configurable settings for LLM parameters (e.g., temperature, thinking time), but it's unclear where to manage them—globally, per model, per agent, or per request.
Currently the Agent class can programmatically set them.

Possible Solutions (Briefly):
Hierarchical Settings: Global → Model → Agent → Chat → Request.
Model-Specific Properties: The model should expose modifiable parameters for users those can then be set per Agent in the Configuration
Chat UI Controls: A settings icon in chat to modify llm settings at request time.

[eneufeld]

Improve Request Settings and Scope Management

Summary

Enhance request settings handling by reusing existing preferences, introducing scope management, and adding client-specific settings.

Tasks

• Reuse current request settings preferences to ensure consistency.
• Introduce a scope object with three optional properties:
  • modelId
  • providerId
  • agentId
• Ensure proper merging of scopes in the following priority order:
  • Agent > Model > Provider
• Implement client settings to retain:
  • Tool calls
  • Thinking state
• Introduce request settings for LLM-specific properties to allow fine-tuned configuration.
• Chat UI should use a chat scope with request settings, where the chat scope takes precedence.

@planger you had local changes on how request settings are handled, we would love to have your input for our suggestion as well as some insights in your planned implementation

[planger]

@eneufeld Looks good to me, thanks!
I've opened the PR for the changes how requests are centralized now, see #15146

[eneufeld]

Fixed via #15092