feat: adding scaling to arx_forecaster

[dsweber2]

Draft of giving the user the option to include population scaling in arx_forecaster. This is something that @brookslogan, @dshemetov, @nmdefries, and I discussed on Wednesday as something that basic users are likely to want easy access to. Defaults to current behavior (i.e. scaled=FALSE).
Example call

jhu <- case_death_rate_subset %>%
  dplyr::filter(time_value >= as.Date("2021-12-01"))
out <- arx_forecaster(
  jhu %>% filter(time_value <= as.Date("2021-12-23")), "death_rate",
  c("case_rate", "death_rate"),
  args_list = arx_args_list(scaled = TRUE, scaling_df = state_census)
)

and the result:

> out$predictions %>%
+   select(-c(forecast_date, .pred_distn)) %>%
+   inner_join(jhu, by = join_by(geo_value, target_date == time_value))
+ # A tibble: 56 × 5
   geo_value   .pred target_date case_rate death_rate
   <chr>       <dbl> <date>          <dbl>      <dbl>
 1 ak        0       2021-12-30       36.4     0.0395
 2 al        0.631   2021-12-30       73.7     0.154 
 3 ar        0.546   2021-12-30       72.0     0.523 
 4 as        0.00727 2021-12-30        0       0     
 5 az        1.38    2021-12-30       39.8     0.635 
 6 ca        5.74    2021-12-30       76.0     0.140 
 7 co        1.11    2021-12-30       93.5     0.462 
 8 ct        0.506   2021-12-30      143.      0.333 
 9 dc        0       2021-12-30      295.      0.0802
10 de        0.521   2021-12-30      119.      0.449 
# ℹ 46 more rows
# ℹ Use `print(n = ...)` to see more rows

Things still to do before this could be merged:

• document the new feature
• provide the same behavior in arx_classifier
• possibly write some tests