Fix!: stop treating dbt schema data types as columns_to_types

[georgesittas]

After changing sushi_dbt as I did in this PR, the following commands run successfully and I can see unused_column in local.duckdb:

$ dbt -t duckdb seed
$ dbt -t duckdb run

Whereas SQLMesh fails to execute the project, because it currently treats the columns in schema.yaml as the source of truth, which does not match dbt's behavior:

$ sqlmesh plan  # Error: In model 'sushi.top_waiters', a description is provided for column 'waiter_id' but it is not a column in the model.

I verified locally (using dbt-duckdb) that setting data_type in schema.yml does not determine the actual DDL type. For example, for a simple model like SELECT 1 AS c, the following entry does not result in a TEXT type for c:

models:
  - name: simple_model
    description: "A simple model with a single column"
    columns:
      - name: c
        description: "A constant value"
        data_type: text

The data_type appears to mostly be relevant when contract is enforced, in which case it acts as an audit. For example, if I add the following config to the above model:

    config:
      contract:
        enforced: true

I see this error at runtime (dbt compile ran fine):

$ dbt run
...
22:52:33    Compilation Error in model simple_model (models/simple_model.sql)
  This model has an enforced contract that failed.
  Please ensure the name, data_type, and number of columns in your contract match the columns in your model's definition.

  | column_name | definition_type | contract_type | mismatch_reason    |
  | ----------- | --------------- | ------------- | ------------------ |
  | c           | INTEGER         | VARCHAR       | data type mismatch |