[HARD] Build LSTM/GRU Deep Learning Model for Forecasting

[ssndb134]

Description:
Design a recurrent neural network capable of forecasting 30 days of ticket demand.

Tasks:

• Create sliding window dataset.
• Implement LSTM/GRU architecture in src/models/lstm_model.py.
• Validation + early stopping.
• Save model to models/lstm.pt.

Acceptance Criteria:

• Must outperform ARIMA baseline.
• No data leakage allowed.