Add specification for new sparse vector query; deprecate specification for text_expansion and weighted_tokens queries

[kderusso]

Adds specification for the new sparse_vector query and deprecates the specifications for text_expansion and weighted_tokens queries.

[l-trotta]

if I got this right every sparse query should have either query_vector or inference_id, the way we model this to have it checked client side is to add /** @variants container */ on top of the class and mark all properties that can always be added with @variant container_property, an example of this is PinnedQuery where you must set either ids or docs

[kderusso]

if I got this right every sparse query should have either query_vector or inference_id, the way we model this to have it checked client side is to add /* @variants container / on top of the class and mark all properties that can always be added with @variant container_property, an example of this is PinnedQuery where you must set either ids or docs

Thanks @l-trotta! I think I got this right?

[l-trotta]

almost! all properties that aren't query_vector or inference_id should be @variant container_property, so not just field. basically this is to assure that either query_vector or inference_id are set, but not both of them, while all the other properties are unaffected by this check.