Bugfix: Pattern match incompleteness for enum was hiding more serious issues

[T-Gro]

Tests are issue reports:
fixes #15597
fixes #14621

What was happening before:

• The first ever incompleteness warning was reported, and later warnings did not go trough counterexample generating phase at all
• This makes sense in reducing the number of warnings, however pattern match incompleteness has (as of now) two different warnings - a generic one and a special one for enums
• Therefore, it is easily possible to have one or the other disabled. With this in action, it could mean that a warning is shadowed by one that is ignored => not good

The approach has been redone to keep:

• The same IL by generating the failure action just once (this is important)
• Reporting each type of warning just once (this is subjective/opinionated: Could be even further imploded to report everything, but would lead to even more warnings and therefore more regressions in error baselines)
• Keep regressions in test suites only to the desired ones (matching the reported bugs)