non-mechanistic model doesn't connect final fit correcly

[sbfnk]

Looking at #695 I noticed that there's a bit of a discontinuity in the non-mechanistic model (or whatever we may want to call it), see also e.g. the vignette plot (not super clear but visible in the infections plot):

https://epiforecasts.io/EpiNow2/dev/articles/estimate_infections_options-nonparametric-1.png

I think this is because there's an issue with how the data is extrapolated to fill future dates in

EpiNow2/R/create.R

Line 122 in 6c6d90c

shifted_reported_cases <- shifted_reported_cases[