[Bug]: grouping with numpy.datetime64 keys gives silently incorrect behavior

[shoyer]

What happened?

All NumPy datetime64 values seem to get mapped to "not a time" (NaT):

import numpy as np
import apache_beam as beam
 
inputs = [(np.datetime64('2000-01-01'), 'first'), (np.datetime64('2000-01-02'), 'second')]
print(inputs | beam.GroupByKey())
# [(numpy.datetime64('NaT'), ['first', 'second'])]

This looks like potentially the same underlying issue as #33020 (incorrect handling of NumPy's int64, which gets mapped to zero)

Issue Priority

Priority: 2 (default / most bugs should be filed as P2)

Issue Components

• Component: Python SDK
• Component: Java SDK
• Component: Go SDK
• Component: Typescript SDK
• Component: IO connector
• Component: Beam YAML
• Component: Beam examples
• Component: Beam playground
• Component: Beam katas
• Component: Website
• Component: Infrastructure
• Component: Spark Runner
• Component: Flink Runner
• Component: Samza Runner
• Component: Twister2 Runner
• Component: Hazelcast Jet Runner
• Component: Google Cloud Dataflow Runner