Add get_feature_names_out() to DataFrameMapper

[pancholi-deep]

Summary

This PR adds optional support for retrieving transformed output feature names
from DataFrameMapper via a new public method, get_feature_names_out().

Motivation

After transformation, DataFrameMapper returns a NumPy array, which makes it
difficult to associate transformed features with their original input columns.
This limits debuggability, interpretability, and downstream analysis in
real-world machine learning pipelines.

Changes

• Track transformed feature names during fitting
• Add get_feature_names_out() public API to DataFrameMapper
• Support transformers that already expose get_feature_names_out
• Provide sensible fallback naming for other transformers
• Add focused tests covering common transformation scenarios
• Document the feature in the README

Backward Compatibility

• Existing behavior is unchanged unless get_feature_names_out() is explicitly called
• No public APIs were modified or removed

Tests

• All existing tests pass
• New tests added for feature-name output behavior