fix(deps): add sentencepiece

Some tokenizers rely on sentencepiece for fast/slow conversions.
Add it to base dependencies so model loading succeeds.

Author: jkoelker

pyproject.toml

@@ -38,6 +38,7 @@ dependencies = [
     "gguf>=0.6.0",
     "kernels>=0.3.0",
     "protobuf>=4.25.0",
+    "sentencepiece>=0.2.0",
     "httpx>=0.25.0",
 ]