feat: infer dbt models from dataset

[betodealmeida]

The CLI currently offers a command to populate dbt exposures — these are charts and dashboards in Superset that were created based on datasets that map back to dbt models.

Currently, the CLI uses extra metadata that is added to datasets when the CLI syncs them from dbt to Superset, eg:

{
  "datasource_type": "table",
  "schema": "public",
  "table_name": "messages_channels",
  ...
  "extra": {
    "resource_type": "model",
    "unique_id": "model.superset_examples.messages_channels",
    "depends_on": "ref('messages_channels')"
  }
}

This makes the process of determining which charts and dashboards read from dbt models easy, since we can just check the extra metadata.

A customer has expressed the need for populating exposures from Preset when they have created the dataset manually. Since the datasets don't have the extra metadata, we need to infer the model references from the schema and table name.

[betodealmeida]

Closes #130.