Add Home Load Support

[markoceri]

Summary

This PR introduces home load monitoring and forecasting — the system now tracks household appliance energy consumption (boiler, EV charger, dishwasher, etc.) and predicts future usage. This allows the mining engine to make smarter start/stop decisions based on expected household demand.

What's New

• Device profiles — Define which appliances are in your home and categorize them (controllable, continuous, seasonal, occasional).
• History tracking — Pulls real consumption data from Home Assistant to build a usage history for each device.
• Energy forecasting — Multiple forecasting strategies (from simple averages to ML-based models like XGBoost) predict how much energy your home will need in the next 1–4 hours.
• Automatic model training — The system trains and evaluates forecast models overnight, automatically promoting the most accurate one.
• Rule engine integration — New fields available in start/stop rules (e.g. home_load.total_forecast.next_2h.avg_power) so you can write conditions like "don't mine if the house will need more than 3 kW in the next hour".
• REST API — Endpoints for managing forecast and history providers.

How It Works (High Level)

1. The system fetches recent power readings from Home Assistant.
2. A trained model forecasts expected consumption for the coming hours.
3. The forecast is injected into the decisional context alongside solar, battery, and grid data.
4. Your start/stop rules can reference predicted home load to decide whether mining should run.

Example Rules

# Stop mining if household demand will be high
- name: high_home_load_stop
  condition: "home_load.total_forecast.next_1h.avg_power > 3000"
  action: stop

# Start mining when household demand is low
- name: low_home_load_start
  condition: "home_load.total_forecast.next_2h.avg_power < 1000"
  action: start