v1.8.4: Fix bundled model loading and storage backend schema

Bug Fixes

1. Dense Model Alias Resolution

• Fixed `e5-small` alias not resolving to bundled `dense_en` folder
• `_parse_model()` now checks `_DENSE_ALIASES` before determining model type
• Pass resolved model name to `DenseEmbedder(model=...)`

2. Storage Backend Schema (Lakebase)

• Pass `mode` in config when creating collection on storage backend
• Mode inferred from tags: `["Ultimate"]` → `mode="ultimate"`
• Lakebase now creates correct columns: `dense_embedding`, `sparse_embedding`, `late_interaction_embedding`

Install

```bash
pip install vectrixdb==1.8.4
```

Usage

```python
from vectrixdb import Vectrix, VectrixDB

lakebase = VectrixDB.with_lakebase(host="...", database="...", user="...", password="...")

db = Vectrix(
"products",
mode="ultimate",
dense_model="e5-small", # Now correctly loads bundled dense_en
sparse_model="bm25",
reranker_model="L12",
late_interaction_model="colbert",
storage_backend=lakebase,
)
```