[Vis Augmenter] Optimize visualization rendering

[ohltyler]

Currently, there is a sequential approach to fetch each augment-vis saved object, execute each plugin expression function, and concatenate results. Ideally, this can be optimized to run in an async way and/or group executions together. Listing some ideas below:

• A refactoring could be done to group all objects of a particular plugin into a single expression function. This would involve refactoring augment-vis objects to represent an association of a plugin resource type to a visualization, rather than a plugin resource to a visualization. And within that single association, could persist the set of plugin resources. When reading and executing, logic could be made to pass all plugin resources to a single plugin expression function which can do batch processing and return all concatenated results at once.
• Process the saved object and expression function executions in an async way. This may involve changing the expression function return values, since right now they are designed to be sequential and build up a final set of results
• Other optimizations of the VisLayer processing (injecting results in source datatable, construction of the final vega spec, etc.)