`flatline_forecaster` `NA` bug

[dsweber2]

If there are NAs at the end of the dataset, somehow flatline_forecaster ends up with a forecast that is repeated 3 times over.

here's constructing a synthetic example that has 2 days of NA values at the end

jhu <- epidatasets::covid_case_death_rates %>%
  dplyr::filter(time_value >= as.Date("2021-11-01"))
geo_values <- jhu$geo_value %>% unique()
one_day_nas <- tibble(
  geo_value = geo_values,
  time_value = as.Date("2022-01-01"),
  case_rate = NA,
  death_rate = runif(length(geo_values))
)
second_day_nas <- one_day_nas %>%
  mutate(time_value = as.Date("2022-01-02"))
jhu_nad <- jhu %>%
  as_tibble() %>%
  bind_rows(one_day_nas, second_day_nas) %>%
  epiprocess::as_epi_df()

with the result

res <- flatline_forecaster(jhu_nad, "case_rate")
+     res$predictions %>% arrange(geo_value)
> # A tibble: 168 × 5
   geo_value .pred .pred_distn forecast_date target_date
   <chr>     <dbl>   <qtls(7)> <date>        <date>     
 1 ak         36.4      [36.4] 2022-01-02    2022-01-09 
 2 ak         36.4      [36.4] 2022-01-02    2022-01-09 
 3 ak         36.4      [36.4] 2022-01-02    2022-01-09 
 4 al         89.9      [89.9] 2022-01-02    2022-01-09 
 5 al         89.9      [89.9] 2022-01-02    2022-01-09 
 6 al         89.9      [89.9] 2022-01-02    2022-01-09 
 7 ar         82.6      [82.6] 2022-01-02    2022-01-09 
 8 ar         82.6      [82.6] 2022-01-02    2022-01-09 
 9 ar         82.6      [82.6] 2022-01-02    2022-01-09 
10 as          0           [0] 2022-01-02    2022-01-09 
# ℹ 158 more rows
# ℹ Use `print(n = ...)` to see more rows