feat: add on-device AI with Ollama integration and comprehensive tests

[rprabhat]

Summary

• Add Ollama client for on-device LLM inference (qwen2.5-coder, nomic-embed-text)
• Add NL-to-SQL translation for natural language queries
• Add document ingestion pipeline (CSV/MD/JSON/TXT chunking)
• Add ModelSettings component for model management
• Add 263 Vitest tests with 46% coverage (app, api, nl-to-sensibleql, components)
• Fix jsdom ESM compatibility by switching to happy-dom

Changes

Rust Backend (Tauri)

• Added src/ollama/client.rs — OllamaClient with check_status, generate_embeddings, chat_completion, pull_model
• Added src/commands/ollama.rs — Tauri commands: ollama_check_status, ollama_pull_model, generate_embeddings, translate_nl_to_sensibleql
• Added src/commands/ingestion.rs — ingest_document, chunk_text, extract_entities_from_chunks

Frontend (SolidJS)

• Added src/components/settings/ModelSettings.tsx — Ollama status, model download with progress
• Updated src/components/chat/ChatView.tsx — LLM-based NL-to-SQL
• Updated src/components/onboarding/ConnectionWizard.tsx — document ingestion pipeline
• Updated src/components/sidebar/Sidebar.tsx — AI Models navigation

Testing

• Added vitest.config.ts with happy-dom (fixes ESM compatibility)
• Created 7 test files: stores, api, nl-to-sensibleql, ModelSettings, ChatView, ConnectionWizard, Sidebar
• 263 tests passing