Batch embedding in pinecone add_texts method (#2657)
Hello there,
I noticed the `add_texts` method took a fair while for the Pinecone
VectorStore. This happened because embeddings were computed one by one.
This PR fixes it by calling `Embeddings.embed_documents` on all texts at
once.
I also took the liberty to harmonize initialization of embeddings to the
Embeddings class.
Cheers,
Olivier
