Fix DataFrame mutation bug by copying data on initialization

[mystic74]

Fixes #205

Summary

This PR fixes the mutation bug described in #205 where the input DataFrame is modified in-memory when include_null=True.

Implemented Fix

We implemented a conditional copy to balance safety and performance. Instead of copying the DataFrame on every instantiation, we only copy when mutation would actually occur:

if self._categorical and self._include_null:
    # Create a copy to avoid mutating the user's original DataFrame
    data = data.copy()
    data[self._categorical] = handle_categorical_nulls(data[self._categorical], self._categorical)
    

Changes

Tests: Added 5 comprehensive regression tests covering numeric/string categorical types and edge cases.
Code: Added a surgical copy to minimize performance impact on large DataFrames.

Verification

All 53 tests pass, including the new regression tests which previously failed.