Take into account accurate times in FOV plotting

[ebuchlin]

The observer for each FOV plot is taken at the mid-time of the observation (see FileMetadata.get_observer(), but the delay between this reference time and the actual time of each exposure can be quite large.

This delay should be taken into account, for the left and right edges of the FOV (last and first slit positions), and in principle also for each point on the top and bottom boundaries. However, one needs to be careful to interpret the WCS headers correctly, especially what they mean in case target locking (following solar rotation) is activated.

[ebuchlin]

This concerns 1 observation; there would be the same issue when plotting the FOVs for several observations over a long duration.

[ebuchlin]

Maybe use SunPy's propagate_with_solar_surface context manager.