Refactor CLI and LLM runtime handling for improved user experience

- Removed deprecated console usage in favor of structured logging functions for error handling and user prompts.
- Enhanced CLI prompts for LLM runtime and preset selection, utilizing menu helpers for better user interaction.
- Updated GPU pool scanning output with improved formatting and error indication for clarity.
- Refactored setup wizard to streamline LLM runtime and preset configuration, ensuring a more intuitive setup process.
- Improved documentation and user feedback in CLI outputs to enhance overall usability.

tests/test_server_plan.py

@@ -31,7 +31,7 @@ def test_prompt_server_plan_with_ranked(monkeypatch, tmp_path):
         "gpu_rent.ux.list_ranked_flavors",
         lambda flavors, cfg: [f1, f2],
     )
-    answers = iter(["2", "200", "n"])
+    answers = iter(["2", "200", "2"])
 
     def ask(msg: str, default: str = "") -> str:
         return next(answers)