Add top k semantic search endpoint

[dgarnitz]

What To Build

• An endpoint add to the api that allows the user to pass in a text-only search term, embed it with a supported model (right now only the 3 OpenAI embedding models), query the vector DB
• This should return the the raw vectors plus the associated metadata. VectorFlow's standard metadata schema includes the source text
• Make the default K 5 or 10 but enable the user to set this
• Add a method that does this to the vectorflow client
• Add a script to the `testing scripts' that servers as an end-to-end test of the whole system
• No reranker is currently needed. If there is demand, this can be a follow up task