feat: enhance CMake configuration and CI workflow for AI features

- Added new build options for agent UI and remote automation in CMake presets, improving modularity.
- Updated CI workflow to enable AI runtime and agent UI during builds, ensuring compatibility with new features.
- Adjusted default Ollama model in scripts and documentation to reflect the lightweight version used in CI.
- Enhanced agent test suite script to support model overrides, improving flexibility for testing scenarios.

docs/public/developer/testing-without-roms.md

@@ -53,6 +53,9 @@ The `agent_test_suite.sh` script now defaults to mock ROM mode:
 
 # Or with Gemini
 ./scripts/agent_test_suite.sh gemini
+
+# Override the Ollama model (CI uses qwen2.5-coder:0.5b)
+OLLAMA_MODEL=qwen2.5-coder:0.5b ./scripts/agent_test_suite.sh ollama
 ```
 
 To use a real ROM instead, edit the script:

docs/public/usage/z3ed-cli.md

@@ -20,7 +20,8 @@ cmake --build --preset mac-ai --target z3ed
 
 The AI features require at least one provider:
 - **Ollama (local)** – install via `brew install ollama`, run `ollama serve`, then set
-  `Z3ED_OLLAMA_MODEL=qwen2.5-coder:7b` (or any supported model).
+  `OLLAMA_MODEL=qwen2.5-coder:0.5b` (the lightweight default used in CI) or any other supported
+  model. Pass `--ai_model "$OLLAMA_MODEL"` on the CLI to override per-run.
 - **Gemini (cloud)** – export `GEMINI_API_KEY` before launching `z3ed`.
 
 If no provider is configured the CLI still works, but agent subcommands will fall back to manual