Fix AnnotationExcluder

[BraisGabin]

#4368 showed that AnnotationExcluder had some issues: false-positives and false-negatives.

This PR implements a FullQualifiedNameGuesser to archive that. It's a guesser so there are some scenarios where we can't know for sure which class we are refering too. But in general it only returns one option. This is a best-effort for the users that doesn't use type-solving. With type-solving we have other implementation that works all the times.

[codecov]

Codecov Report

Merging #4518 (fab3a88) into main (5176e41) will increase coverage by 0.30%.
The diff coverage is 98.18%.

@@             Coverage Diff              @@
##               main    #4518      +/-   ##
============================================
+ Coverage     84.11%   84.41%   +0.30%     
- Complexity     3312     3323      +11     
============================================
  Files           477      479       +2     
  Lines         10920    11129     +209     
  Branches       2029     2037       +8     
============================================
+ Hits           9185     9395     +210     
  Misses          700      700              
+ Partials       1035     1034       -1     

Impacted Files	Coverage Δ
...ub/detekt/psi/internal/FullQualifiedNameGuesser.kt	89.28% <89.28%> (ø)
...gitlab/arturbosch/detekt/api/AnnotationExcluder.kt	94.73% <93.75%> (+13.48%)	⬆️
...ithub/detekt/psi/internal/KotlinNoImportClasses.kt	100.00% <100.00%> (ø)
...in/io/gitlab/arturbosch/detekt/rules/style/Junk.kt	75.00% <0.00%> (-25.00%)	⬇️
...n/io/github/detekt/report/html/HtmlOutputReport.kt	95.50% <0.00%> (-0.05%)	⬇️
...otlin/io/gitlab/arturbosch/detekt/core/TaskPool.kt	100.00% <0.00%> (ø)
...rturbosch/detekt/rules/style/TrailingWhitespace.kt	96.29% <0.00%> (+0.29%)	⬆️
...lab/arturbosch/detekt/rules/style/MaxLineLength.kt	95.00% <0.00%> (+0.40%)	⬆️
...in/kotlin/io/github/detekt/report/xml/XmlEscape.kt	65.97% <0.00%> (+1.58%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5176e41...fab3a88. Read the comment docs.

[marschwar]

Non blocking comment: I hate parameterized tests like this. That is why I split it into 2 tests. But I know there are more schools of thought here ;) But that is something for our style guide. How many parameters may a test have before it is considered unreadable.

[BraisGabin]

I don't like this kind of tests either. I would prefer to have way less. But they were needed. There were really random cases that were not covered. And while I was implementing this I broke some of them while fixing others.

And I think that it's better to have all the cases together because they are in blocks of 5. Maybe I could add a new line between each "block of tests" to make it easier to read. But I can reverse the change and split them in two if you think it's better.

[cortinico]

A couple of things here:

1. I would move it in a separate file.
2. Is it copied from somewhere in the Kotlin repo? If so, please provide a reference so we can keep it up to date easily.

[BraisGabin]

Both of them done. Thanks to force the point 2. I was missing lots of classes. I couldn't find a list with this information so I created a script to extract it from the documentation of detekt.

Basically I download the documentation and parse the html. And I extracted the list of packages to parse from the documentation (all explained in the comment)