Tests failing on CRAN for `show_values()` when API is down

[mjwestgate]

Another round of automated checks has failed on CRAN, meaning we need to update by 2023-03-11 to keep {galah} on CRAN. Issue is:

  ── Failure ('test-international-Sweden.R:60'): search_all(fields) 
works for Sweden ──
   nrow(x) is not more than 1. Difference: -1
   ── Error ('test-international-Sweden.R:73'): show_values works for 
Sweden ──────
   Error in `if (!(field %in% show_all_fields()$id)) {
       bullets <- c("Unknown field detected.", i = "Search for the valid 
name of a desired field with `search_fields()`.")
       abort(bullets, call = caller_env())
   }`: argument is of length zero
   Backtrace:
       ▆
    1. ├─vcr::use_cassette(...) at test-international-Sweden.R:73:2
    2. │ └─cassette$call_block(...)
    3. └─galah::show_values(search_fields("basisOfRecord")) at 
test-international-Sweden.R:74:4
    4.   ├─base::do.call(paste0("show_values_", type), args)
    5.   └─galah:::show_values_field(field = NULL)

What appears to be happening here is:

• Sweden API goes down (this time; the principle is the same for other systems)
• The call search_fields("basisOfRecord") |> show_values() is not cached correctly by var::use_cassette for some reason
• show_all_values() doesn't error (correct behaviour), but also doesn't return useful information
• show_values() fails noisily, despite not calling an API at all

Some potential solutions:

• make show_values() fail more quietly when incorrect information is provided
• make the failing tests conditional on a non-NULL result from show_all(values)
• cache more show_all(fields) calls (currently only done for ALA and GBIF)
• run more tests without internet to catch odd behaviour
• make more liberal use of testthat::skip_on_cran