[Feature] Modify expansion of `BaseRetriever` to match `BaseGenerator` for multimodal considerations

[nerdai]

Tasks:

• Update fed_rag.base.retriever to use Query and Context
• Update fed_rag.base.retriever to return a new encoder type that flexible accepts separate encoders for text and image (audio and video later)
• Update fed_rag.retrievers.hf_sentence_transformer to be able to take in Query and Context inputs — ignoring the other modalities, namely image (audio and video later).