Remove atomic setter transformers

[eupp]

Remove separate bytecode transformers which tracked methods like .set, .compareAndSet, etc, of VarHandle, AFU, and Unsafe APIs, to inform the ObjectTracker about changes in object graph topology (registerObjectLink listener).

Instead, modify ManagedStrategy::beforeMethodCall to parse the method name and arguments to extract relevant information and pass it into ObjectTracker.

Also, it looks like the new approach is more robust and handles more cases, which were not handled previously (e.g. atomic array APIs). As such, there are changes in some representation tests --- the reported traces now include more events, which were previously incorrectly classified as local object accesses.

[ndkoval]

@eupp, do we have tests checking that the local object detection works correctly with updates via methods like .set, .compareAndSet, etc, of VarHandle, AFU, and Unsafe APIs

[eupp]

@eupp, do we have tests checking that the local object detection works correctly with updates via methods like .set, .compareAndSet, etc, of VarHandle, AFU, and Unsafe APIs

Apparently, there are tests for AFU, Unsafe, and VarHandle APIs, but it looks like there are no tests for Atomics and Atomic Arrays. I can add them.

lincheck/src/jvm/test/org/jetbrains/kotlinx/lincheck_test/transformation/LocalObjectsTests.kt

Line 139 in c4745c8

class AtomicUpdaterLocalObjectsTest {

lincheck/src/jvm/test/org/jetbrains/kotlinx/lincheck_test/transformation/LocalObjectsTests.kt

Line 231 in c4745c8

class UnsafeLocalObjectsTest {

lincheck/src/jvm/test-jdk11/org/jetbrains/kotlinx/lincheck_test/transformation/VarHandleLocalObjectsTest.kt

Line 29 in c4745c8

class VarHandleLocalObjectsTest {

[eupp]

I have added new tests to check that local object detection works correctly with AtomicReference and AtomicReferenceArray classes. Tests for other APIs (AFU, VarHandle, Unsafe) were already there.