Docs: Add how-to article for running models (API and local)

[kosar]

Goal

Add a short, practical how-to article that helps users choose and run models (LLMs) for their agents: metered API providers and local/self-hosted options.

Context

We already reference docs/MODEL-CHOICE.md and mention Ollama in the README. Users need a clear path to (1) use API-backed models and (2) run models locally (e.g. Ollama) as an alternative to metered APIs. This how-to should set them up to make that choice and get it working.

What the how-to should cover

• Using metered API models: Where to configure API keys and model names (e.g. OpenAI, Moonshot, Anthropic), and how the agent uses them.
• Running models locally: How to "roll up your sleeves" and run a local model (e.g. Ollama) that can be used in place of metered API providers.
• Practical note: Include a neutral note that local models can be much slower than API-based ones and may cause agent responses to take minutes rather than seconds—so users can set expectations and choose accordingly.

Format

Keep it concise: how-to style, not a long reference. No bias toward either approach; the article should support both. It can live under docs/ (e.g. docs/HOWTO-MODELS.md) and be linked from the main README or MODEL-CHOICE where appropriate.