⚡ Performance: np.append in loop creates O(n²) memory allocation

## Description

In `setup_db.py` line 18:
```python
embedding_db = np.empty((0, embedding_size), dtype=np.float32)
for page in tqdm(pdf):
content = page.get_text()
embedding = embedding_model.encode([content])
embedding_db = np.append(embedding_db, embedding, axis=0) # O(n²)
```

`np.append` copies the entire array each iteration, resulting in O(n²) time and memory. For large PDFs this becomes extremely slow.

## Suggested Fix
Collect embeddings in a list, then stack once:
```python
embeddings = []
for page in tqdm(pdf):
content = page.get_text()
embedding = embedding_model.encode([content])
embeddings.append(embedding)
embedding_db = np.vstack(embeddings)
```