froll could support window size 0

[jangorecki]

Previously we were prohibiting n smaller than 1.
After recent PR a revdep started to fail and Ben spotted that n argument validation was not raising an error for adaptive rolling funs using n<=0.

frollmean(1:2, n=c(0,0), adaptive=TRUE)
#> NaN NaN
frollmean(1:2, n=c(-1,0), adaptive=TRUE)
#> NA NaN
frollmean(1:2, n=0)
#> Error in froll(fun = "mean", x = x, n = n, fill = fill, algo = algo, align = align,  :
#>   n must be positive integer values (> 0)

Reviewing a revdep, bidask package, it appeared it was desired behavior.
Considering base R works for zero length input I believe we should also provide such support

mean(numeric())
# [1] NaN