Add more backends

[nleroy917]

Right now, supernova has been built/optimized for a single pipeline and/or pathway:

HF --> [nova-embed] --> S3 --> [nova-dist] --> Qdrant

But, to be more useful, I think we should support more backends -- particularly in:

1. Where to dump embeddings after generated, (more buckets? can we abstract "bucket" storage? i.e. R2, Azure, etc)
2. Embedding loading destinations (Add Milvus, Elastic, OpenSearch, Pinecone, etc)

[nleroy917]

I addressed no. 1 in 269183c