Onboarding: add optional Ollama (local LLM) setup step

[victorlucss]

Idea

During onboarding, offer an optional step to set up Ollama so users can enable local LLM cleanup right away (grammar/punctuation/filler removal, tone presets) without sending data to the cloud.

Proposed flow (onboarding, optional step before "Try it")

• Title: "Smarter cleanup (optional)" — explain LLM cleanup + that Ollama runs locally/private.
• Detect if Ollama is installed/running (e.g. probe http://localhost:11434/api/tags).
  • If running: offer to pull a small model (e.g. llama3.2) and enable cleanup (set llmEnabled=true, llmProvider=ollama).
  • If not: a "Get Ollama" button that opens https://ollama.com/download (or brew install ollama), with a re-check.
• Skippable; cleanup stays off if skipped.

Notes / open questions

• We can't silently install Ollama (separate app); best we can do is detect + link + pull a model via its API.
• Reuse the existing model-download progress pattern for the Ollama model pull.
• Should pair with surfacing this in Preferences → LLM too.

Deferred — tracking. Lower priority than the cross-platform work (#3).