Check `field_is_present` always passes for CSV/Parquet files

[jschoedl]

For CSV and Parquet files that are obtained over DuckDB (S3, GCS, Azure and local environment), the check field_is_present always passes, independent of whether the column is present in the CSV or not.

Cause: In create_view_with_schema_union(...), we create an empty table based on the datacontract and later insert the actual data, e.g. from a CSV file. If a column is not present in the CSV file, it will still be in the checked table (but filled with NULLs).

Possible solutions:

1. Remove the field_is_present check (for non-required fields). However, it is useful to check for field presence as discussed in Support for historical data validation causes error in CSV files without headers #1018.
2. Fix the check and enforce field presence in the header even for non-required fields. This would break, e.g., test_csv_optional_field_missing_from_old_data() in test_test_schema_evolution.py, which assumes that a CSV-file that misses a non-required field should pass all tests.