Provide a deep schema validation in the preprocessing step

So far we fully rely on pandas csv parser to do its job.

• Define a pydantic schema to be assessed.
• Check that nullable / empty array fields are consistenlty parsed as null / False fields thorughout the pipeline.