UnusedDetector: add missing RECORD label

[mbien]

fixes #8928

[lahodaj]

Looks good. The only enhancement I could see is adding a test.

[mbien]

@jlahoda added tests for the hint.

There is one curiosity though: initially I wanted to add test cases to the UnusedDetectorTest, since it would be the most direct way to validate the functionality.

e.g:

    @Test
    public void testNoUnusedPackagePrivateClass() throws Exception {
        performTest(
                "test/Test.java",
                """
                package test;
                class Test {
                }
                """
        );
    }

However, it would fail since returns package private top level classes/interfaces/records.. etc as unused.

Testcase: testUnusedPackagePrivateClass(org.netbeans.modules.java.editor.base.semantic.UnusedDetectorTest):	FAILED
expected:<[]> but was:<[2:<init>:NOT_USED]>

I debugged through it and it correctly marks the class declaration as used, but later sees an synthetic constructor and marks that as unused. I believe what is happening afterwards is that since synthetic items have no line numbers, it maps the unused line to the class declaration so that it looks as if it would think that the class should be marked as unused. (constructor name == class name too so the error line becomes indistinguishable)

So why does the hint work?
the hint then drops the item at

netbeans/java/java.hints/src/org/netbeans/modules/java/hints/bugs/Unused.java

Line 70 in 8e2bc77

if (ud.unusedElementPath().getLeaf() != ctx.getPath().getLeaf()) {

That is why the junit test works as expected for the hint, while the UnusedDetector produces technically wrong output in that case.

However, changing this is not required to resolve this issue and it would be unnecessarily risky for the release. We should probably fix this at some point though.

I think this can get in as is (+ the added hint tests)

[lahodaj]

FWIW, retroactivelly (sorry), still looks good.
Re the synthetic problem - we might look into solving that (basically, no compute the unused status for synthetic declarations), but the tests ought to be OK for now, I think.

Thanks!

[mbien]

@lahodaj yep, would be good to properly fix UnusedDetector, but that would be something for NB 29 I think.