feat: Allow transformations on writes to output list of entities

[franciscojavierarceo]

What this PR does / why we need it:

This PR refactors the feature transformation logic in the FeatureStore class to improve the efficiency and clarity of the code. Additionally, it adds a new unit test class TestOnlineWritesWithTransform to test the transformation logic specifically for PDF inputs. The changes aim to enhance the overall performance and maintainability of the feature transformation process.

• sdk/python/feast/feature_store.py

  • Refactored the feature transformation logic for singleton feature views to handle row-wise transformations more efficiently.
  • Replaced the previous df.apply approach with an iterative row processing method to improve performance and clarity.

• sdk/python/tests/unit/online_store/test_online_writes.py

  • Added imports for necessary modules and types.
  • Introduced a new test class TestOnlineWritesWithTransform with a test method test_transform_on_write_pdf to verify the transformation logic for PDF input data.
  • Defined new entities and request source for PDF input data.
  • Implemented an on_demand_feature_view with a transformation function transform_pdf_on_write_view to handle PDF input transformations.

• sdk/python/tests/unit/test_on_demand_python_transformation.py

  • Updated the test_docling_transform method to handle multiple input samples.
  • Adjusted the docling_transform_docs function to process and verify multiple input samples.
  • Ensured that the transformation logic correctly writes multiple chunks to the online store and verifies the stored data.

Which issue(s) this PR fixes:

#5210

Misc

N/A

[ntkathole]

for k, v in output.items():
        if isinstance(v, list):
            transformed_rows[k].extend(v)
        else:
            transformed_rows[k].append(v)

[franciscojavierarceo]

I'll do that in a follow up PR, don't want to have to rerun integration tests 😭

[ntkathole]

It's non-blocking 👍

[ntkathole]

Looks good 👍

[shuchu]

Lgtm