Add `sort` and `ascending` arguments to `value_counts`

[ajpotts]

Summary

Add pandas-compatible sort and ascending arguments to Arkouda-backed value_counts implementations to align behavior with pandas.Series.value_counts.

Motivation

In pandas, value_counts sorts by frequency by default (sort=True, ascending=False).
Current Arkouda-backed implementations return results in server-dependent order, which:

• diverges from pandas expectations,
• makes results harder to reason about,
• forces tests to ignore ordering.

Adding explicit sorting controls improves API parity while preserving a path to efficient implementations.

Pandas reference behavior

Series.value_counts(
    normalize=False,
    sort=True,
    ascending=False,
    bins=None,
    dropna=True,
)

• sort=True: sort by count
• ascending=False: highest counts first
• sort=False: preserve order of appearance (after dropna)

Proposed API

Add arguments to all Arkouda-backed arrays that implement value_counts:

def value_counts(
    self,
    dropna: bool = True,
    sort: bool = True,
    ascending: bool = False,
):
    ...

Semantics

• Default behavior matches pandas (sort=True, ascending=False)
• Sorting is performed client-side on the (typically small) (keys, counts) result
• When sort=False, document that ordering reflects server return order (not guaranteed to be first-seen order)

Scope

• ArkoudaArray.value_counts
• ArkoudaStringArray.value_counts
• ArkoudaCategoricalArray.value_counts
• Docstrings and user-facing documentation
• Unit tests

Testing

• Verify default ordering matches pandas for representative inputs
• Verify sort=False preserves unsorted output (order-insensitive comparison)
• Verify interaction with dropna

Non-goals

• No requirement to implement stable "first-seen" ordering when sort=False
• No server-side sorting changes in this ticket

Acceptance criteria

• API matches pandas defaults
• Tests cover sorted and unsorted cases
• Documentation clearly states ordering guarantees