[ML] Deprecate the text_expansion task type and replace with sparse_embedding

[davidkyle]

The ELSER model and other sparse embedding models are configured in the ML trained models API with task type text_expansion. sparse_embedding is now the preferred way to describe the model task. The text_expansion query associated with the text_expansion task type is already deprecated and replaced by the sparse_vector query, similarly the task type should be updated to sparse_embedding.

[elasticsearchmachine]

Pinging @elastic/ml-core (Team:ML)