feat(install): auto-install Ollama, start server, and pull models

[blackaxgit]

Summary

• clx install now auto-installs Ollama via brew install ollama if missing
• Auto-starts ollama serve if binary exists but server is not running
• Auto-pulls required models (qwen3:1.7b, qwen3-embedding:0.6b) if Ollama is running
• Graceful fallbacks at every step: warns and continues if any step fails

Install Flow

clx install
  -> Check Ollama binary (not found? brew install ollama)
  -> Check server running (not running? start ollama serve, poll up to 10s)
  -> Check models (missing? POST /api/pull for each, 600s timeout)
  -> Create directories, config, rules, prompts
  -> Copy binaries to ~/.clx/bin/
  -> Initialize SQLite database
  -> Configure Claude Code hooks + MCP in settings.json
  -> Inject CLX section into CLAUDE.md

Test plan

• Fresh install on clean HOME: 20 items installed, 0 errors
• Health after install: 8 pass, 1 warn
• Uninstall (no purge): hooks/MCP removed, data preserved
• Reinstall after uninstall: idempotent, 0 errors
• Uninstall --purge: ~/.clx/ deleted completely
• Double uninstall edge case: graceful no-op, exit 0
• Install after purge: full fresh install works
• Automated test suite: 162 tests pass, 0 clippy warnings
• No Homebrew scenario: warns and continues (L0-only mode)
• Ollama server timeout: warns and skips model pulling

Generated with Claude Code