Cause an error when non-existing columns are selected in `collect_observation_record(var=...)`

The following code returns an error because "AGEU" is not included in meta$data_observation. It would be better to raise an error message for this situation.

library(metalite)
library(forestly)
adsl <- forestly_adsl
adae <- forestly_adae
adsl$TRTA <- factor(adsl$TRTA, levels = c("Placebo", "Xanomeline Low Dose"))
adae$TRTA <- factor(adae$TRTA, levels = c("Placebo", "Xanomeline Low Dose"))


meta <- meta_adam(population = adsl, observation = adae) |>
  define_plan(plan = plan(analysis = "ae_forest_plot",
                          population = "apat",
                          observation = "oapat",
                          parameter = "any-AE"           
  )) |>
  define_population(name = "apat",
                    group = "TRTA",
                    subset = SAFFL == "Y" ,
                    label = "All Participants as Treated Population",
                    var = c("USUBJID", "TRTA", "RACE", "SEX", "AGE", "AGEU")) |>
  define_observation(name = "oapat",
                     group = "TRTA",
                     subset = SAFFL == "Y" ,
                     label = "All Participants as Treated Population") |>
  define_analysis(name = "ae_forest_plot",
                  label = "Interactive AE Forest Plot") |>
  define_parameter(name = "any-AE",
                   subset = SAFFL == "Y",
                   label = "Any AEs",
                   var = "AEDECOD", soc = "AEBODSYS", term1 = "", term2 = "") |>
  meta_build()

collect_observation_record(meta, "apat", "oapat", "any-AE", var = c("USUBJID", "TRTA", **"AGEU"**))

Start at the collect_observation_record() entry point and trace how the var argument is checked against the observation metadata. Reproduce the example with AGEU, then add coverage for a missing column and verify that the function raises a clear error identifying the unavailable variable.