Reference datetime changes by ~seconds across disp_s1 runs

[scottstanie]

From @gracebato

we found out that the recent products (v06) have varying reference time. So for example, in this list of produced data, you can see that, for example, the reference date 20160927 for frameID F08882 have different times in the filename:
OPERA_L3_DISP-S1_IW_F08882_VV_20160927
T002623Z_*
T002624Z_*
T002628Z_*
T002630Z_*
T002633Z_*
T002635Z_*
T002636Z_*
T002637Z_*
T002644Z_*
T002647Z_*

https://github.com/nasa/opera-sds-pge/blob/main/src/opera/pge/disp_s1/disp_s1_pge.py#L285-L287 the name comes from looking in what we stick in disp_metadata['identification']['reference_datetime']

disp_s1_pge.py

        reference_date_time = disp_metadata['identification']['reference_datetime']
        reference_date_time = datetime.datetime.strptime(reference_date_time, "%Y-%m-%d %H:%M:%S.%f")
        reference_date_time = f"{get_time_for_filename(reference_date_time)}Z"

I believe this means we must not be populating the compressed SLC metadata correctly, or we are not passing it through across multiple batch runs