Please raise error or warning in `epidata()` on some 0-row output cases

[brookslogan]

library(epidatr)
#> ! epidatr cache is being used (set env var EPIDATR_USE_CACHE=FALSE if not
#>   intended).
#> ℹ The cache directory is /home/fullname/.cache/R/epidatr.
#> ℹ The cache will be cleared after 14 days and will be pruned if it exceeds 4096
#>   MB.
#> ℹ The log of cache transactions is stored at
#>   /home/<username>/.cache/R/epidatr/logfile.txt.
# querying a non-existent geo_type simply yields 0 rows with no warning:
smoothed_pct_ed_visits_covid_not <- epidata("nssp", "smoothed_pct_ed_visits_covid", "hsa")
nrow(smoothed_pct_ed_visits_covid_not)
#> [1] 0
# vs. querying an actually-supported geo_type:
smoothed_pct_ed_visits_covid_hsa <- epidata("nssp", "smoothed_pct_ed_visits_covid", "hsa_nci")
nrow(smoothed_pct_ed_visits_covid_hsa)
#> [1] 139185

^{Created on 2026-08-14 with reprex v2.1.1}

It'd be nice if the first query could raise an error and list the correctly-"spelled", applicable geo types for the source.

However, some geo types are not comprehensively supported. For example, some HSAs don't have NSSP data in CA (e.g., Alameda Contra Costa, which I believe should be 766). That's not just a simple check against epidata_meta("nssp")$nssp$geo_types. And one could get into the same situation when specifying time ranges as well. I think we discussed what to do in these cases in v4, and it seems that we ended up keeping a warning when returning 0 rows? That should be easy enough, though not as helpful as the much-more-complicated explanations of CA NSSP availability or outages /inactive systems.