Improves Unit testing capabilities.

Enabler for: https://github.com/fishtown-analytics/dbt/issues/2354 Discussed here: https://discourse.getdbt.com/t/testing-with-fixed-data-set/564/8

Related: https://discourse.getdbt.com/t/state-of-testing-in-dbt/1778

Description

It didn't seem to be a good(relatively speaking) method for testing models in isolation. The best was introducing branching conditions in the model based on global settings which at most allowed for 1 test / model.

This PR introduces a new helper-method model_sql() (the name is debatable, in previous discussions it was called cte) which returns the compiled SQL code of a model but under a different context (allows one to overrides var/ref/source ). Models can be tested without changes outside test code and multiple tests can be build for the same model.

I think think approach is a cleaner approach to (unit) testing than what has been proposed so far.

1. Pure dbt (Jinja/SQL)
2. No need for custom macros.
3. No additional plugins (if this pr gets merged)
4. “Full" control per test
5. No changes/branching/flags required in production models (!)
6. Supports all data sizes.

Example use case:

Model models/addder.sql

    SELECT 
      *,
      a+b + {{ var ("base", "0") }} AS sum
    FROM {{ ref('source_data') }}

Test tests/adder_0001.sql

{% set model_to_test = model_sql(
    'model.adder',  {
      'var': {
        'base': '1'
      },
      'ref': {
        'source_data': 'mocked_source_data'
      }})
%}

WITH
mocked_source_data AS (
  SELECT
    *
  FROM
  -- bigquery construct; same effect as union all select ....
  UNNEST([
     struct(1 as a,  2 as b,  4 as expectedSum),
     struct(2 as a,  2 as b,  5 as expectedSum),
     struct(0 as a,  0 as b,  1 as expectedSum),
     struct(1 as a, -2 as b, 0 as expectedSum)
   ])
),

final AS (
  {{ model_to_test }}
)

-- Some SQL assertion based on the specific mocked data.
SELECT * FROM final WHERE sum != expectedSum

_I haven't been writing python for some 6 years and and this is the first time interacting the the DBT core, as such the PR is not yet complete. I would like to get some feedback before investing more time into this. No point in "polishing a turd".

If the maintainer(s) think it holds value i don't mind ticking all the boxes below (or hand it over to someone else)_

Checklist

• I have signed the CLA
• I have run this code in development and it appears to resolve the stated issue
• This PR includes tests, or tests are not required/relevant for this PR
• I have updated the CHANGELOG.md and added information about my change to the "dbt next" section.