Implement partner demand forecasting module

Build a TypeScript module that predicts future item requests from a partner organization so banks can proactively stock inventory.

forecastDemand(
  partnerId: string,
  itemId: string,
  horizon: 30 | 60 | 90  // days ahead to forecast
): Promise<ForecastResult>

// ForecastResult: {
//   predicted: { month: string, units: number, confidence: 'high' | 'medium' | 'low' }[],
//   seasonalityDetected: boolean,
//   dataPointsUsed: number
// }

Algorithm:

1. Fetch monthly distribution totals for the partner+item from the database (last 24 months max)
2. Apply simple exponential smoothing (α = 0.3): S_t = α * x_t + (1 - α) * S_{t-1}
3. Detect seasonality: if the coefficient of variation across same-month values across years > 0.2, set seasonalityDetected: true
4. Confidence based on data available: 'high' = 12+ months, 'medium' = 6-11, 'low' = under 6
5. Return one predicted entry per month for the forecast horizon

Write unit tests for 3 scenarios:

• Steady demand (should forecast near the historical average)
• Seasonal pattern with summer spikes
• Fewer than 6 data points (should return 'low' confidence and still produce a forecast)