Duration subdivision KGO updates after resolving forecast period and cell method issues

[bayliffe]

KGO for duration subdivision plugin updated such that the forecast periods reflect the end point and ranges of the new shorter periods. The inclusion of duplicate cell methods has also been prevented.

Example

Consider kgo_daymask.nc The subdivided cube has 2 shorter 3-hour periods.

First period

forecast_period             1220400 seconds, bound=(1209600, 1220400) seconds
forecast_reference_time     2024-12-03 00:00:00
time                        2024-12-17 03:00:00, bound=(2024-12-17 00:00:00, 2024-12-17 03:00:00)

We can check that the forecast period is correct, this measuring between the forecast reference time and the time point.

(cube[0].coord("time").cell(0).point - cube[0].coord("forecast_reference_time").cell(0).point).total_seconds()
>>> 1220400.0

Second period

We can do the same for the second period.

forecast_period             1231200 seconds, bound=(1220400, 1231200) seconds
forecast_reference_time     2024-12-03 00:00:00
time                        2024-12-17 06:00:00, bound=(2024-12-17 03:00:00, 2024-12-17 06:00:00)

For the point difference we get:

(cube[1].coord("time").cell(0).point - cube[1].coord("forecast_reference_time").cell(0).point).total_seconds()
>>> 1231200.0

which again, is correct. We can also check that the lower bound of the forecast period range aligns with the lower bound of the forecast time:

(cube[1].coord("time").cell(0).bound[0] - cube[1].coord("forecast_reference_time").cell(0).point).total_seconds()
>>> 1220400.0

and indeed it does.