Closes #5406: Series accessor groupby

[ajpotts]

Summary

This PR adds a groupby() method to the ArkoudaSeriesAccessor,
enabling pandas-style grouping directly from an Arkouda-backed
pd.Series:

s = pd.Series([80, 443, 80]).ak.to_ak()
g = s.ak.groupby()
keys, counts = g.size()

This eliminates the need for users to access internal _data attributes
and provides a clean, abstraction-safe API.

---

Motivation

Previously, users had to write:

ak.GroupBy(s.values._data).size()

Accessing ._data breaks encapsulation and bypasses the ExtensionArray
abstraction layer.

This PR introduces:

s.ak.groupby()

which: - Preserves the pandas accessor pattern - Avoids NumPy
materialization - Returns a proper arkouda.pandas.groupbyclass.GroupBy
object - Maintains zero-copy behavior

---

Implementation Details

Changes

arkouda/pandas/extension/_series_accessor.py

• Import GroupBy from arkouda.pandas.groupbyclass
• Add groupby(self) -> GroupBy method to ArkoudaSeriesAccessor
• Validate that the Series is Arkouda-backed
• Safely extract the underlying Arkouda array (_data)
• Return GroupBy(akcol)

The method raises:

• TypeError if the Series is not Arkouda-backed
• TypeError if the underlying _data attribute is missing

This ensures correctness and defensive behavior.

---

Tests

Added tests under:

tests/pandas/extension/series_accessor.py

Coverage includes:

• ✔ Raises when Series is not Arkouda-backed
• ✔ Returns ak.GroupBy
• ✔ g.size() matches pandas value_counts().sort_index()
• ✔ Raises when _data is unavailable

This validates both happy-path and failure cases.

---

API Behavior

Example

>>> s = pd.Series([80, 443, 80]).ak.to_ak()
>>> g = s.ak.groupby()
>>> keys, counts = g.size()

Equivalent to:

ak.GroupBy(s.array._data).size()

But without leaking internal implementation details.

---

Design Notes

• No materialization to NumPy occurs.
• This maintains the zero-copy Arkouda-backed semantics.
• Aligns with future plans to expand pandas-style groupby support.

---

Future Extensions (Optional)

This lays the groundwork for:

• g.sum()
• g.min() / g.max()
• Returning pandas Series directly from groupby aggregations
• Full DataFrame groupby parity

Closes #5406: Series accessor groupby

[1RyanK]

Looks good!