Add Ollama service to docker-compose with GPU/CPU auto-detection

[berardifra]

Summary

Add the ollama container to the IntelOwl Docker Compose stack as part of the GSoC 2026 LLM Chatbot integration.

Tasks

• Add ollama service to docker/default.yml (or override, TBD pending mentor input)
• GPU passthrough config (NVIDIA runtime)
• CPU fallback for environments without GPU
• Init script that pulls the default model on first boot
• Environment variables for LLM config in docker/env_file_app_template

Notes

Part of W1: Infrastructure & Scaffolding. Depends on mentor confirmation on whether Ollama should be optional (override file) or required (base compose).

[mlodic]

The ollama service should be considered optional. Please use an override and also modify the start script to support this new container