Make tests robust to exact phrasing & language of base condition messages

[MichaelChirico]

Closes #4789

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.53%. Comparing base (7fbc314) to head (45bfbd0).

❗ Current head 45bfbd0 differs from pull request most recent head 466a539. Consider uploading reports for the commit 466a539 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6029      +/-   ##
==========================================
- Coverage   97.53%   97.53%   -0.01%     
==========================================
  Files          80       80              
  Lines       14916    14915       -1     
==========================================
- Hits        14548    14547       -1     
  Misses        368      368              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[tdhock]

this seems reasonable. not sure how often base R would make a change to its messages?
at least this refactoring makes it clear in our test file which error messages are from us, and which are from base R.

[MichaelChirico]

not sure how often base R would make a change to its messages?

not terribly often, but it's no fun to get kicked off CRAN for it when it does happen, as I just dealt with in a patch release for {lintr}.

it's also a better testing practice in principle: only directly test the code we own, though as seen it's a bit more tedious to design the tests & slightly cumbersome to avoid regression. on balance I think avoiding CRAN issues is worth these overheads which can themselves be mitigated.