[Discuss] Separate out data source configurations from existing visualization types

[joshuarrrr]

Today, one of the implicit assumptions in OpenSearch Dashboards visualization and dashboard implementations is that the data source for any visualization is likely the result of a query to OpenSearch. As we move toward support for multiple datasources, we need to make sure our visualization abstractions clearly separate visualization options and configurations from the source of the data.

We already have some existing abstractions to assist with this - for example most visualization rendering expression functions take OpenSearchDashboardsDatatable as their input. But we need to make sure the visualization types themselves are data source agnostic. In addition, it should be much easier for developers to "bring their own data" to a visualization - including doing so without understanding or knowing about the expression plugin.

We may even need to add support for other types of data formats or structures, in order to efficiently render data results from PPL queries (as currently implemented in the Observability plugin.

[seraphjiang]

@joshuarrrr @kavilla @mengweieric @ashwin-pc
are we achieved this goal in recent import ppl/sql to discover app project?