feat: Add support for Pinecone vector store

[cvauclair]

Vector Store Integration Request

Pinecone is a popular cloud based, fully managed vector database designed specifically for RAG applications.

Resources

The Pinecone API provides a user-friendly way to insert and query embeddings. This should make the integration pretty straighforward.

Note that users will require a Pinecone API key to be able to use Pinecone. Moreover, the responsibility of creating a collection and an index on that collection is outside the scope of Rig (the integration should assume that the user has already setup Pinecone correctly).

Interestingly, Pinecone is a pure vector store database, in the sense that it does not allow users to store documents, only vectors (and optional metadata for filtering). Performing vector search only returns the IDs of the closest vectors and not full documents.

Another interesting observation is that Pinecone has an endpoint to generate embeddings. This means that Pinecone could potentially also be integrated as an embedding model provider.