Implement first database/storage layer #38

Goal

Implement the first DuckDB-based storage layer for ARGUS based on the storage research.

Why

ARGUS needs persistent local storage to move beyond temporary API requests and in-memory analytics. A first storage layer makes it possible to save normalized historical market data, read it back for analytics and prepare the project for later storage-first workflows.

This ticket focuses only on the database/storage foundation. Workflow orchestration, API fallback and service refactoring are handled in separate sub-issues.

Scope

• Add DuckDB as the first storage dependency

• Create the first minimal DuckDB schema based on the storage research:

  • data_sources
  • instruments
  • price_bars

• Store normalized internal model data, not raw API response structures

• Add a small storage/repository module for:

  • initializing the database
  • inserting/upserting data sources
  • inserting/upserting instruments
  • saving price bars
  • reading price bars by instrument, source and date range

• Keep storage logic separate from API clients, UI code and metric functions

• Add tests for schema creation, saving and reading historical data

• Document basic setup and usage

Out of scope

• Storage-first/read-through workflow
• API fallback when data is missing
• Refactoring TimeSeriesService into MarketDataService
• Refactoring existing services
• GUI integration
• Analytics/reporting integration
• Cache strategy

Acceptance criteria

• DuckDB is added as the selected first storage technology
• A minimal DuckDB schema exists for data_sources, instruments and price_bars
• The schema follows the storage research in docs/research-databases-and-storage.md
• Historical FX or market data can be saved using the internal data model
• Stored price bars can be read back by instrument, source and date range
• FX rates are stored as close values in price_bars
• Storage code does not depend directly on raw API response structures
• Storage logic is separated from API clients, UI code and metric functions
• Basic tests cover database initialization, saving data and reading data
• README or docs explain how to use the storage layer