Test split_duration_records_vectorized thorougly

[izzet]

This pull request updates the split_duration_records_vectorized function to consistently use the standard int64 dtype for the COL_TIME_RANGE column, and introduces a comprehensive new test suite to verify the function’s correctness across a variety of scenarios, including edge cases and Dask dataframe integration.

Data type consistency:

• Changed the dtype of the COL_TIME_RANGE column from 'uint64[pyarrow]' to standard 'int64' in both the shortcut and main vectorization paths within split_duration_records_vectorized in dfanalyzer/analysis_utils.py. This ensures compatibility and avoids potential issues with pyarrow extension types. [1] [2]

Testing improvements:

• Added a new test module tests/test_analysis_utils.py that provides extensive coverage for split_duration_records_vectorized, including:
  • Parameterized tests for different time granularities and resolutions.
  • Validation of correct chunk splitting, handling of zero durations, and preservation of count/size totals.
  • Edge case tests for durations not evenly divisible by granularity, large granularity, and microsecond-level events.
  • Integration tests for Dask dataframe partitioning, ensuring correct behavior across different partitioning schemes and with/without zero-duration rows.