change ~all uses of window_width_n to window_width in vignette

[mikeblazanin]

using window_width is just better practice, because of NA handling and consistency across runs w/ dift gap times

demonstrate the best practice across all vignettes

[mikeblazanin]

On second thought, just tell people to think about which to use and point out that both are available.

It's true that _n handles NA's perhaps slightly better, but that only applies for NA's in the raw data (not those induced by transformation e.g. trans_y in calc_deriv), which is unlikely to be very common

And it's actually ambiguous whether it's better to average across different numbers of points or across different time windows without running the simulations yourself and checking mathematically which introduces less bias

[mikeblazanin]

done in v0.12.1.9000