Setup for Wrangler SQL Execution Support

[harshdeeppruthi]

This PR is responsible for the following:

1. Adding the initial setup for the SQL Execution support for Wrangler Directives using CDAP's relational API
2. Adding a UI toggle to enable this SQL Execution of directives

This includes implementation of sql support of drop directive as an example. Once this is merged, the SQL support for other directives can be extended in a similar way

[minurajeeve]

You could add a description for this method similar to getCountMetrics()

[albertshau]

Looking just at this class, it seems like every existing directive would become a no-op in SQL, which would be incorrect. It only makes sense when seeing the isSQLSupported() method in the RelationalTransform interface, and that it defaults to false.

I think it would be cleaner if the Directive interface did not extend the RelationalTransform interface, with only directives that do support SQL (like Drop) implementing it. Then the isSQLSupported() method is not needed, Wrangler just checks if the directive is an instance of RelationTransform. This way it is impossible to implement a directive where the transform() and isSQLSupported() methods don't match up.

[albertshau]

nit: add a newline after this line

[albertshau]

It's better to name this something like 'executionMode' and have an internal Enum for it. This is more extensible than a boolean toggle for SQL or no SQL, as it allows us to add new modes in the future if needed. (As a side note, when a property is meant to be a toggle, it should be of Boolean type, not String type).

For example, we may eventually want to add an 'auto' mode that lets the engine choose, or maybe some BigQuery specific SQL mode or something like that.

[albertshau]

JEXL is a technical term that many users won't understand. It's also not technically correct, as many directives are direct java implementations that don't use jexl scripting. Not sure what would be a better term though. Native? Default?