Closes #5413: bug in pd.array when dtype is ak.bigint

[ajpotts]

Summary

Fixes a bug where constructing a pandas ExtensionArray from an
existing Arkouda bigint pdarray using:

pd.array(a, dtype="ak.bigint")

would incorrectly attempt a server-side cast:

cast<bigint,object,1>

which fails because Arkouda does not support casting bigint to
object.

---

Root Cause

Inside ArkoudaArray._from_sequence, when pandas passed an
ArkoudaBigintDtype, the implementation converted it to its NumPy
equivalent (object).

This caused:

1. ak.array(existing_bigint_pdarray, dtype=object)
2. Which triggered ak.cast(..., object)
3. Resulting in an unsupported backend command: cast<bigint,object,1>

---

Fix

Special-case ArkoudaBigintDtype in _from_sequence so that:

• "bigint" is passed directly to ak.array
• No conversion to NumPy object occurs
• No unnecessary server-side cast is attempted

All other Arkouda dtypes continue to use their NumPy equivalents as
before.

---

Test Coverage

Adds regression test:

test_pd_array_construct_from_ak_bigint_pdarray_explicit_dtype_does_not_cast_to_object

This test verifies:

• Construction succeeds without server error
• The resulting array is Arkouda-backed
• Values are preserved
• The dtype remains bigint (allowing alias tolerance)

This would have failed prior to this fix with a RuntimeError.

---

Impact

• Fixes explicit ak.bigint dtype construction
• Prevents invalid backend cast commands
• Improves pandas interoperability for Arkouda bigint arrays
• No breaking changes

Closes #5413: bug in pd.array when dtype is ak.bigint