do rolling window offset correction based on file_datetime

[japhir]

in stead of doing it based on the number of measurements before/after the current one (currently a window of 200 around the measurement), do it based on the file_datetime itself so that if there was a holiday or gap in measurements, they don't get added.

Use R package slider for this, especially the slider::hop function.

[japhir]

some more old notes:
research rollapply vs time: make offset_correction(_wrapper) work on file_datetime in stead of row so that skipped samples are not taken into account

offset_correction_wrapper

• perhaps switch to slide from https://davisvaughan.github.io/slider/articles/slider.html?
• use .before = lubridate::days(1) and .after = lubridate::days(1). leave .complete = FALSE default, so that it also calculates it at the edges of the windows.

in that case it would be slide_index.