Update the solar angle calculations to use per-line times.

[simonrp84]

Feature Request

At present, the SunZenithCorrectorBase computes the solar zenith angle based upon the start_time of a given scene - computed on line 73 of modifiers/geometry.py:

coszen = xr.DataArray(cos_zen(vis.attrs["start_time"], lons, lats), dims=['y', 'x'], coords=coords)

This is okay, but leads to some inaccuracies. For SEVIRI, pixels at the end (north) of the scan are sensed 12 minutes after the start_time, and for some sensors the difference in time between sensing and given time may be even larger. I have not checked, but this would be a significant issue if done for polar satellites providing full-orbit granules, as the solar position may be out-of-date by up to 90 minutes!

Describe the solution you'd like
It would be a useful improvement if we calculate the solar angle based on a per-scanline time (where available) or at the least an average of the scene start and end times, as this would minimise error across the scene.

Describe any changes to existing user workflow
No changes required by users, although this would slightly affect (positively) the output from composites that perform sun zenith correction.

[sfinkens]

Good idea! Would be nice if this was available as an utility function. Here's how it is done for SEVIRI in level1c4pps, maybe that helps: https://github.com/foua-pps/level1c4pps/blob/58519bc6b3fc7f78234484fff57f88491b2630a7/level1c4pps/seviri2pps_lib.py#L176