make the use of `loader_file_format` parameter consistent

[rudolfix]

Background
The loader_file_format can be configured on the pipeline level and this is fine. But technically it is a normalize setting and we should be able to configure it there as well.

Also not all data types are supported by all formats on all destinations. We should detect those and fail early.

Tasks

1. • add loader_file_format to normalize configuration and use it to set the file format. If the explicit value is passed from the pipeline, it has the precedence.
2. • add the exceptions to destination capabilities where we declare unsupported types per file format. Take the current cases from particular destination code (ie Bigquery and Snowflake have plenty of exceptions)
3. • detect such exceptions in normalize (preferably in worker: before updating a schema check if any type is on exception list and raise exception with an explanation)
4. • the behavior above must be tested

[mankala29]

Hi! I can take this on. Let me know if it’s still open or if there’s any context I should be aware of before I start.

[YinkaMetrics]

I can take a scoped pass at the normalize config part of this: adding a supported normalize.loader_file_format setting and keeping the explicit pipeline/resource value as higher precedence. I will check for existing behavior first and keep the PR focused with tests, rather than trying to bundle the destination-specific unsupported-type matrix into the same change.

[AshwabhB]

Picking up tasks 2 to 4 (task 1 looks covered by #4151). Starting with task 3: wiring the existing verify_supported_data_types capability check into the normalize worker so unsupported data type and file format combinations fail during normalize instead of at load time. This reuses the current load time logic, so destinations that already declare exceptions (bigquery, redshift, athena) get the earlier check for free. Snowflake's use_decfloat/parquet exception (task 2) and tests will follow. Anything you'd want done differently on the normalize side before I open a PR?

[AshwabhB]

Opened #4172 for task 3 (the normalize-stage check). Went ahead with the approach described above; happy to adjust in review. Tasks 2 and 4 to follow.