Fix binary64 ndarray encoding overflow for signed zero

[awf]

encode_ndarray(format_info_binary64, ...) raises OverflowError: Python int too large to convert to C long when NumPy resolves the signed-zero np.where operands. The negative-zero code is 2**63, outside the signed int64 range. This also affects arrays with no zero values and empty arrays because NumPy still resolves the operands for an empty selection.

Cast both zero code points to np.uint64 before passing them to np.where, matching the output array's dtype. Add regression coverage for signed zeros, mixed two-dimensional inputs, inputs without zeros, and empty arrays. The tests compare exact float64 bit patterns and verify output shape and dtype.

Fixes #63.

[orlitzky]

There was probably no doubt in your mind, but I can confirm that this does fix my tests.

[Copilot]

Copilot review overview

🟢 Approval recommended

The fix directly addresses the reported overflow and has focused coverage for all described cases.

Review effort: Lite
Findings: None

What changed in this PR

Fixes binary64 ndarray encoding overflow for signed zero by using uint64 zero code points and adding regression tests.

Changes:

• Cast zero code points to np.uint64 before np.where.
• Add coverage for signed zeros, multidimensional, nonzero, and empty arrays.

File	Description
src/​gfloat/​encode_ndarray.py	Prevents signed-zero code conversion overflow.
test/​test_encode.py	Adds regression tests for affected inputs.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[awf]

There was probably no doubt in your mind, but I can confirm that this does fix my tests.

Thanks!