[Feature] Implement on-device Python model execution

[b-re-w]

On-device Python model execution is not yet implemented

• The app should run models locally using Python within the mobile environment, but this capability remains on the backlog due to limited development time.

Add on-device Python model execution

1. Set up a Python runtime within the app’s sandboxed environment (e.g., via a Python interpreter library or bundled modules).
2. Implement a lightweight inference pipeline that can load models and execute them directly on the device.
3. Expose an API or service layer for initiating local inference requests from the rest of the app.
4. Add configuration options for model paths and resource limits (CPU/GPU usage, memory).
5. Provide documentation and samples demonstrating how to deploy and execute on-device models.

Multiple provider (server selection) support is missing

• Users should be able to choose from various servers—such as Gemini, Claude, or other Gemstone instances—to run their models, but this feature is currently lower priority and not yet available.

Implement multiple provider server selection

1. Design a provider registry that can list available servers and their capabilities (e.g., Gemini, Claude, Gemstone peers).
2. Implement server selection logic and corresponding UI elements for choosing a provider.
3. Integrate authentication and API request handling for each supported provider.
4. Create fallbacks or error handling for unreachable providers or mismatched model capabilities.
5. Document the server configuration process and how to add new providers in the future.