chore(dbt Core): Bypass metric evaluation if final SQL is provided

[Vitor-Avila]

The CLI currently uses MetricFlow to retrieve the syntax for a metric. This current process has two issues:

• The mf command can take a few minutes for each metric (causing a full sync to take a long time)
• The mf command might return a SQL query with un-necessary JOINs that is not compatible with dataset metrics in Superset.

To workaround this issue, this PR adds support for specifying the metric SQL syntax that should be used in Superset directly in metric.meta.superset.expression. In this case, since the metric won't be evaluated it's also required to specify the metric.meta.superset.model (so that the CLI knows which dataset to associate the metric):

metrics:
  - name: test_distinct
    description: test
    type: simple  # Will be ignored in this case
    label: Test
    type_params:
      measure: distinct_orders  # Will also be ignored
    meta:
      superset:
        expression: COUNT(DISTINCT order_id)
        model: model.vec_sales.vehicle_sales