Implement `stacksmith ai` with a local Ollama provider

## Summary
`ai` is a stub. Implement module generation that follows the project's conventions, starting with a local Ollama provider (no cloud keys required).

## Tasks
- [ ] Add a provider abstraction (`ollama` first)
- [ ] Read `stacksmith.config.json` for context (stack, paths)
- [ ] Generate files into the existing project with a confirmation step
- [ ] Document `OLLAMA_HOST` / model env vars

## Files
- `src/commands/ai.ts`
- `README.md` (.env docs)

Start in `src/commands/ai.ts` since the issue points to it as the implementation target, and locate how existing commands parse input, read config, and confirm actions. Check existing provider-style patterns in the codebase, then add the Ollama provider flow using `stacksmith.config.json` values for stack and paths. Update `README.md` (`.env` docs) with `OLLAMA_HOST` and model environment variable guidance. Done means `stacksmith ai` can generate files into the current project after confirmation using the local Ollama path and docs are complete.

Escrita pelo modelo de indexação a partir do texto da issue.