Fix!: Type inference for BQ engine adapter

[VaggelisD]

Fixes #3978

In the BigQuery engine adapter, type inference for the fields happens through to_standard_sql(); However, it seems like the types that do not belong to stanard SQL are not handled properly (e.g RANGE or JSON, see 4951ff9) and are instead marked as TYPE_KIND_UNSPECIFIED.

This PR adds a fallback for these types and instead uses field.field_type, as was used before:

MODEL (name vaggelis.test);

SELECT 
 RANGE('01-01-1900'::DATE, '01-01-1902'::DATE) AS my_range, 
 JSON '{"id": 10}' AS foo;

>>> bq_adapter.columns("vaggelis.test");
{'my_range': DataType(
  this=Type.RANGE,
  expressions=[
    DataType(this=Type.DATE, nested=False)],
  nested=True), 'foo': DataType(this=Type.JSON, nested=False)}

[georgesittas]

LGTM, let's test these changes.