Dataset loading script method does not work with .pyc file

### Describe the bug

The huggingface dataset library specifically looks for ‘.py’ file while loading the dataset using loading script approach and it does not work with ‘.pyc’ file.
While deploying in production, it becomes an issue when we are restricted to use only .pyc files. Is there any work around for this ?

### Steps to reproduce the bug

1. Create a dataset loading script to read the custom data.
2. compile the code to make sure that .pyc file is created
3. Delete the loading script and re-run the code. Usually, python should make use of complied .pyc files. However, in this case, the dataset library errors out with the message that it's unable to find the data loader loading script.

### Expected behavior

The code should make use of .pyc file and run without any error.

### Environment info

NA