chore: move embedding deps to RAG tool where they are needed (#1210)

`EMBEDDING_DEPS` were wrongly associated with `vector_io` providers.
They are needed by
https://github.com/meta-llama/llama-stack/blob/main/llama_stack/providers/utils/memory/vector_store.py#L142
and related code and is used by the RAG tool and as such should only be
needed by the `inline::rag-runtime` provider.

llama_stack/templates/vllm-gpu/build.yaml

@@ -4,6 +4,7 @@ distribution_spec:
   providers:
     inference:
     - inline::vllm
+    - inline::sentence-transformers
     vector_io:
     - inline::faiss
     - remote::chromadb

llama_stack/templates/vllm-gpu/vllm.py

@@ -20,7 +20,7 @@ from llama_stack.templates.template import (
 
 def get_distribution_template() -> DistributionTemplate:
     providers = {
-        "inference": ["inline::vllm"],
+        "inference": ["inline::vllm", "inline::sentence-transformers"],
         "vector_io": ["inline::faiss", "remote::chromadb", "remote::pgvector"],
         "safety": ["inline::llama-guard"],
         "agents": ["inline::meta-reference"],