Summary

Closes #5829.

Fix NumPy datetime64 arrays containing NaT being serialized incorrectly instead of producing JSON null.

Changes

• Normalize missing datetime values to None in the standard JSON encoder.
• Fix the orjson serialization path to handle NaT correctly without disabling the fast path for unaffected data.
• Add regression tests covering datetime units (D, us, ns), multidimensional arrays, both JSON engines, and pandas Series round trips.
• Update CHANGELOG.md.

Validation

Tested locally on Windows with Python 3.13.14, Plotly 7.1.0, pandas 3.0.6, and NumPy 2.5.4.

• 330 passed, 1 skipped — tests/test_io and tests/test_optional/test_utils
• Ruff formatting check passed for all three modified Python files.
• Full Ruff lint has pre-existing findings; it is not reported as passing.

AI assistance

ChatGPT assisted with investigation, implementation, and regression tests. The reported validation results were obtained from a local Plotly source checkout.