[ENHANCEMENT] Integrate with template query and ext improvements in ML commons

[ohltyler]

There are few enhancements made in core OpenSearch and in ML commons that can help smooth the search experience. Updates in the presets and general UI should be done to leverage these backend updates. Some ideas:

1. For template query, this can eliminate the query overriding required in the ML processors in the default use cases. Rather than having a separate preset query defined in the ML processors, that query can be defined directly within the template clause. The ML processor preset can then be updated to just parse out what fields to parse as input, and where to store in the output (e.g., parsing some input field to generate embeddings for, and store the generated embeddings in some query field)
2. For ext improvements, the RAG use case can be updated to store the ML search response processors LLM response under the ext field. And, explore some enhancements in the search response UI to easily show the ext response.

[ohltyler]

Relevant PRs:
opensearch-project/OpenSearch#16818
opensearch-project/ml-commons#3284

Template query may not be backported to 2.x, so integrating with this feature for 2.x is on hold.
ext improvement is still not merged, with only one day remaining until code freeze, so integrating with this feature for 2.x is also on hold.

Updated label to 3.0 due to the delays.