Sparse Embedding Integration

Integrate sparse embeddings (BM25/SPLADE) with dense embeddings for hybrid search.

## Current Dense Setup
{{current_setup}}

## Sparse Model Options
{{sparse_options}}

## Integration Requirements
{{integration_requirements}}

Implement hybrid integration:

```python
class HybridEmbedder:
    def __init__(self, dense_model: str, sparse_model: str):
        pass
    
    def encode_dense(self, text: str) -> np.ndarray:
        """Generate dense embedding"""
        pass
    
    def encode_sparse(self, text: str) -> Dict[int, float]:
        """Generate sparse embedding (term -> weight)"""
        pass
    
    def encode_hybrid(self, text: str) -> HybridEmbedding:
        """Generate both representations"""
        pass

class HybridIndex:
    def search(self, query: str, alpha: float, top_k: int) -> List[Result]:
        """Combined search with score fusion"""
        pass
```

Include:
- SPLADE model integration
- Score normalization
- Alpha tuning methodology
- Storage optimization