Wire presets from rooms.settings.yaml into the CLI wizard

[rosspeili]

Problem Statement

rooms.settings.yaml supports a presets map (e.g. local-ollama, openai with litellm_model and optional api_key_env), and resolve_preset_model() exists in rooms/settings.py, but the interactive CLI never offers presets. Users must type LiteLLM model strings manually even when presets are already defined in their settings file.

Proposed Solution

• In the custom-agent wizard and orchestrator setup, offer an optional step: choose a preset by name (from settings.presets) or enter a model string manually.
• When a preset includes api_key_env, reuse the existing wizard flow to prompt for that env var (same as non-ollama/ models today).
• Surface preset names in docs (docs/EXAMPLES.md) with a short example block in rooms.settings.example.yaml.
• Add tests: preset resolution + CLI smoke (mocked prompts) where a preset is selected.

Alternatives Considered

• Document presets as “copy-paste model strings only” — rejected; duplicates config and misses the point of feat: externalize default LiteLLM models via YAML settings file #26.
• Auto-apply a default preset without asking — rejected; too surprising for users who customize models per session.

Priority

Medium 🟡

Additional Context

Follow-up to merged PR #28. Presets are parsed and validated today but not wired in cli.py. Keeps settings as the single place for “local Ollama vs cloud” profiles.