Update default Embeddings parameters

[davidmezzetti]

Make the following parameter updates:

• keyword: Allow boolean (sets to BM25 like current) and string (sets to a scoring method)
• sparse: new parameter that enables sparse vector scoring. When set to True it will use a default sparse vector model, otherwise this should be set to a model path
• dense: new parameter that is an alias for path. True sets the default dense vector model, otherwise this should be set to a model path.
• hybrid: True enables BM25 + vector search (like it does today). String supports hybrid indexes with a scoring method or sparse vector path

With this new pattern, embeddings can be created as follows.

from txtai import Embeddings

embeddings = Embeddings(keyword=True)
embeddings = Embeddings(keyword="bm25")

embeddings = Embeddings(sparse=True)
embeddings = Embeddings(sparse="prithivida/Splade_PP_en_v2")

embeddings = Embeddings(dense=True)
embeddings = Embeddings(dense="sentence-transformers/all-MiniLM-L6-v2")

embeddings = Embeddings(sparse=True, dense=True)
embeddings = Embeddings(hybrid=True)
embeddings = Embeddings(hybrid="prithivida/Splade_PP_en_v2")

This change will be fully backwards compatible.