Loading 600K features from a parquet file

### Context
Given the sample files from the [OpenShift AI tutorial - Fraud detection example](https://docs.redhat.com/en/documentation/red_hat_openshift_ai_self-managed/2-latest/html/openshift_ai_tutorial_-_fraud_detection_example/index), [repo link](https://github.com/rh-aiservices-bu/fraud-detection), including 600,000 rows modelling a `fraud detection` dataset.

### Objective
I want to model the same training dataset as a Feast offline store using a `FileSource` and then fetch all the historical features at once.

See all the details to replicate the issue here: [https://github.com/dmartinol/feast_issue_600K](https://github.com/dmartinol/feast_issue_600K)