fix: Image Feature in Datasets Library Fails to Handle bytearray Objects from Spark DataFrames (#7517)

[giraffacarp]

Task

Support bytes-like objects (bytes and bytearray) in Features classes

Description

The Features classes only accept bytes objects for binary data, but not bytearray. This leads to errors when using IterableDataset.from_spark() with Spark DataFrames as they contain bytearray objects, even though both bytes and bytearray are valid bytes-like objects in Python.

Changes

• Updated Features classes to accept both bytes and bytearray types for binary data fields.

Reasoning

• bytes and bytearray serve the same purpose for binary data, with the only difference being mutability.
• Modifying the Spark iterator to convert bytearray to bytes would be a workaround, not a true fix. I think the correct solution is to accept all bytes-like objects as input.
• This approach is more robust and future-proof since Python 3.12+ provides a standard way to check for buffer protocol.

Testing

• Added tests to cover bytearray inputs for image features.

Related Issues

• Fixes: Image Feature in Datasets Library Fails to Handle bytearray Objects from Spark DataFrames #7517

[giraffacarp]

@lhoestq let me know if you prefer to change the spark iterator so it outputs bytes