Refactor redundant code in epi_slide, epi_slide_opt

[brookslogan]

There's redundant/reimplemented logic for padding and completion, maybe time filtering, etc. Also epi_slide implementation seems quite hard to follow through all its pieces (or to swap out / refactor out this duplicate logic), and epi_slide_opt isn't all that friendly either. Let's try eliminating some overlapping code and making things easier to read by unifying what is possible from the slide_one_group/epikey-type functions. E.g., we may want some conversions from the output of as_time_slide_computation, and from data.table and slider slide functions, to a common "simple" "slider"/"hopper" function that will take a padded&completed input time series and add output columns to it. Then the padding&completion interface and output filtering can be shared between epi_slide and epi_slide_opt.