Fix pandas 3.0 KeyError in Enum.encode() for Series with non-integer index

[MaxGhenis]

Summary

Fixes #427

In pandas 3.0, string columns use StringDtype by default. When a pandas Series has a non-integer index (e.g., string index), array[0] does label-based lookup (looking for key "0") instead of positional access, causing KeyError: 0.

This broke policyengine-us when enabling pandas 3.0 support - specifically the county variable tests failed with:

policyengine_core/enums/enum.py:69: in encode
    if len(array) > 0 and isinstance(array[0], Enum):
                                      ^^^^^^^^
E   KeyError: 0

Changes

• Modified Enum.encode() to use .iloc[0] for pandas Series to ensure positional access
• Added two tests for encoding pandas Series containing Enum items:
  • test_enum_encode_pandas_series_with_enum_items - basic Series with default index
  • test_enum_encode_pandas_series_with_string_index - Series with string index (the failing case)

Test plan

• New tests pass locally
• All existing enum tests pass
• CI passes

Related

• policyengine-us pandas 3.0 PR (blocked by this): Add pandas 3.0 compatibility tests policyengine-us#7233

🤖 Generated with Claude Code