`epix_slide` fails on `yearmonth`s

[brookslogan]

suppressPackageStartupMessages({
library(dplyr)
library(tsibble)
library(epiprocess)
})

tibble(geo_value = 1, time_value = yearmonth(1:5), version = yearmonth(6), value = 1:5) %>%
  as_epi_archive() %>%
  epix_slide(~ .x)
#> Error in UseMethod("full_seq"): no applicable method for 'full_seq' applied to an object of class "c('yearmonth', 'vctrs_vctr')"

tibble(geo_value = 1, time_value = yearmonth(1:5), version = yearmonth(6), value = 1:5) %>%
  as_epi_archive() %>%
  epix_slide(~ .x, .versions = yearmonth(6))
#> Warning in max(geo_value): no non-missing arguments to max; returning -Inf
#> # A tibble: 0 × 4
#> # ℹ 4 variables: version <mth>, geo_value <dbl>, time_value <mth>, value <int>

^{Created on 2025-03-28 with reprex v2.1.1}

The first case seems self-explanatory; we rely on full_seq when it doesn't have an impl for yearmonths. The second case's weird max(geo_value) warning is coming from guess_geo_type on an empty epi_df, but the concerning thing is why there's an empty edf to begin with, and that's because <yearmonth> - Inf is an NA yearmonth, leading to NA filter conditions in epix_as_of.