Improved verbosity of data.table:::test() function for mismatched expected warnings

[lucaslarson25]

Closes #7092.

Previously, the data.table:::test() function would print out expected warnings separately:

data.table:::test(1, {warning('a'); 2}, 2, warning=c('a', 'b'))
# Test 1 produced 1 warnings but expected 2
# Expected: a
# Observed: a
#  Test 1 produced 1 warnings but expected 2
# Expected: b
# Observed: a

This was caused by using paste() with the array of expected warnings. To fix this, we collapse these arrays using a new function align_messages.
This behavior was caused by multiple expected warnings, or multiple observed warnings in the test function.
Now the test function will output multiple expected and/or observed warnings correctly:

data.table:::test(1, {warning('a'); 2}, 2, warning=c('a', 'b'))
# Test 1 produced 1 warnings but expected 2
# Expected: a
#           b
# Observed: a

Also, created new tests in inst/tests/tests.Rraw file as well as updated the NEWS.md file.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.04%. Comparing base (8364344) to head (d8cd21f).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #7706   +/-   ##
=======================================
  Coverage   99.04%   99.04%           
=======================================
  Files          87       87           
  Lines       17037    17037           
=======================================
  Hits        16874    16874           
  Misses        163      163           

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[aitap]

Very good.

[aitap]

This is interesting. I don't think test data.table tested its test() function before. Now we can have tests purposefully failing tests so that the outer test can succeed (and the coverage can be increased, removing the # nocov start ... # nocov end comments).

Maybe we need a separate test file for this?

[jan-swissre]

this comment is relevant:

data.table/R/test.data.table.R

Line 450 in 8364344

# num=0 to escape global failure tracking so we can test behaviour of test function itself: test(1.1, test(0, TRUE, FALSE), FALSE, output="1 element mismatch")

tests in this PR seems to follow that