Task 1.2 (b) Move Development Code related to the D2 layer

[xandie985]

D2 Layer Implementation for Time Series Dataset

Key Changes

1. Architecture

   • Split into modules:
     • time_series_d2.py: Model-ready processing (TSDataModule)
   • Modularized data handling to separate D2 logic from D1, improving maintainability and clarity.

2. D2 Layer (TSDataModule)

   • PyTorch Lightning integration
   • Sliding-window generation with NaN-aware indexing (preserves NaNs for downstream validation/model training)
   • Flexible train/val/test splits

3. Integration with D1 Layer (MultiSourceTSDataSet)

   • Multi-source CSV loading, encoding/normalization, memory-efficient & pre-loaded modes
   • Compatibility tests ensuring D2 correctly consumes D1 outputs

4. Code Quality

   • Docstrings, type hints, error handling, logging

5. Testing

   • Unit tests for D2 layer
   • Tests covering D1→D2 integration

⚠️ Blocked by PR #16: Please merge PR #16 to main before continuing work on this PR.