Wrap DBT models/tests in `{% model %}...{% endmodel %} blocks

[beckjake]

Feature

Wrap DBT models/tests in {% model %}...{% endmodel %} blocks

Feature description

When we read in a sql file that is a model, we should optionally wrap it in a {% model %} block. Then we should run dbt's top-level jinja block parser over it and make sure the output matches the contents of the sql file, and either raise or log a warning.

Model blocks should probably be named using the basename of the sql file they were found in just for coherency, but it doesn't really matter.

This should be behind a (hidden?) flag to dbt compile/run.

We should do the same when generating tests - after we build the string, wrap it in a block, extract that block, and make sure it matches.

Who will this benefit?

This will help us shake out any bugs around block parsing.