fix(app): show setup wizard when configured Ollama model is missing

[comppaz]

Summary

Fixes #44

When Ollama is already installed but the configured chat model (e.g., llama3.2) isn't downloaded, the app now properly shows the setup wizard to download the required models.

Changes

• App.tsx: Check if the configured model is actually available before skipping the setup wizard
• SetupWizard.tsx:
  • Check for chat model (not just embedding) when wizard runs
  • Show user-friendly download status instead of raw layer hashes (e.g., "Downloading model..." instead of "pulling dde5aa3fc5ff")

Test plan

• Have Ollama installed and running, but without llama3.2 model
• Launch Think app with think_ai_setup_complete already set in localStorage
• Verify the setup wizard appears and downloads the missing model
• Verify friendly status messages are shown during download
• Verify chat works after setup completes