Standardize source data to input/output transform modals

[ohltyler]

Description

This PR standardizes the generated source input to consistently match exactly the way the ML processors will transform the inputs. For full_response_path=false/true, both are displayed exactly how the ML processor will parse it. The one caveat is, there is still work to be done on the search response side to handle many-to-one/one-to-one toggling better.

Also has other small fixes:

• handles when input or output is missing from the model interface (both are not required)
• handles empty search pipeline edge case when simulating source output
• triggers touched when toggling boolean field - before, changing these values was not triggering save to be available.
• removes the invalid check of bracket notation not available for the input/output map configurations
• adds loading logic when fetching to prevent overloading potentially expensive calls (simulation on ingest/search pipelines can be using ML credits & have large latency)
• fixes bug of showing multi-predict selector when input map is empty

Demo video, showing the updated input transforms and output transform source data for ingest/search req. Extra data, arrays, etc. have all been removed to consistently show the exact input JSON to the ML processor inputs, including when full_response_path is true/false

screen-capture.19.webm

Check List

• Commits are signed per the DCO using --signoff

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[saimedhi]

LGTM