TimeIntervals.get_duration / get_starting_time return NaN with NaN start/stop times

[bendichter]

What happened?

TimeIntervals.get_duration (added in #2146) uses np.max on stop_time data. Stop times are allowed to be NaN (e.g. for ongoing/unbounded intervals), and np.max propagates NaN, so get_duration returns nan whenever any stop time is NaN. The same issue applies to get_starting_time, which uses np.min on start_time.

Expected behavior

• get_starting_time should ignore NaNs (use np.nanmin), and return None if the table is empty or all start times are NaN.
• get_duration should ignore NaNs (use np.nanmin/np.nanmax), with these edge cases:
  • All start times and all stop times are NaN → return NaN.
  • All stop times are NaN but valid start times exist → return the span of the start times (latest start − earliest start).
  • Empty table → return None.

Steps to reproduce

import numpy as np
from pynwb.epoch import TimeIntervals

ti = TimeIntervals(name="intervals")
ti.add_interval(start_time=0.0, stop_time=1.0)
ti.add_interval(start_time=2.0, stop_time=np.nan)  # ongoing interval

print(ti.get_duration())  # -> nan, expected 2.0 (latest valid stop is 1.0; here all-NaN-stop fallback not triggered)

Operating system

N/A

Python version

N/A

Code of Conduct

• I agree to follow this project's Code of Conduct
• Have you ensured this change was not already reported?