Fix the output type of Padding op

[sararb]

Goals ⚽

• The Padding operation currently converts the values of all input columns to float64. However, we want to maintain the original dtypes (essential for embedding lookup of categorical features, for example).
• The existing padding operation does not take the padding_value into consideration, as it always uses zero-padding.

Implementation Details 🚧

• Updated the padding function pad_put_zeros to preserve the original column type.
• Modified pad_put_zeros to consider the padding_value
• Revised the docstrings for the Padding class.

Testing Details 🔍

• Added a unit test for the data loader with ragged embeddings and the transform: Padding() >> Embeddings().

[karlhigley]

Thanks for the fix!